remote sensing

Review
Industry- and Academic-Based Trends in Pavement Roughness
Inspection Technologies over the Past Five Decades:
A Critical Review
Ali Fares * and Tarek Zayed

Department of Building and Real Estate, The Hong Kong Polytechnic University, Hong Kong;
[email protected]
* Correspondence: [email protected]

Abstract: Roughness is widely used as a primary measure of pavement condition. It is also the key
indicator of the riding quality and serviceability of roads. The high demand for roughness data has
bolstered the evolution of roughness measurement techniques. This study systematically investigated
the various trends in pavement roughness measurement techniques within the industry and research
community in the past five decades. In this study, the Scopus and TRID databases were utilized.
In industry, it was revealed that laser inertial profilers prevailed over response-type methods that
were popular until the 1990s. Three-dimensional triangulation is increasingly used in the automated
systems developed and used by major vendors in the USA, Canada, and Australia. Among the
research community, a boom of research focusing on roughness measurement has been evident in
the past few years. The increasing interest in exploring new measurement methods has been fueled
by crowdsourcing, the effort to develop cheaper techniques, and the growing demand for collecting
roughness data by new industries. The use of crowdsourcing tools, unmanned aerial vehicles (UAVs),
and synthetic aperture radar (SAR) images is expected to receive increasing attention from the research
community. However, the use of 3D systems is likely to continue gaining momentum in the industry.

Keywords: road monitoring and maintenance; roughness; IRI (international roughness index);
Citation: Fares, A.; Zayed, T.
profiler; 3D imaging systems; satellite; review; industry practice
Industry- and Academic-Based
Trends in Pavement Roughness
Inspection Technologies over the Past
Five Decades: A Critical Review.
Remote Sens. 2023, 15, 2941.
1. Introduction
https://doi.org/10.3390/rs15112941 For several decades, pavement condition surveys have been conducted by highway
agencies to collect a variety of pavement condition data. However, roughness (sometimes
Academic Editors: Riccardo Roncella,
Massimo Losa and Nicholas
identified as smoothness, rideability, or riding comfort) has been the most commonly
Fiorentini
measured, particularly at the network level [1,2]. Pierce and Stolte [2] found that virtually
100% of agencies in the United States (USA) collect roughness data at the network level.
Received: 10 April 2023 Roughness is used as one of the primary indicators of pavement conditions [3]. Additionally,
Revised: 19 May 2023 it is utilized by many agencies to select the treatment type of pavement sections [4–8].
Accepted: 1 June 2023 Pavement roughness is particularly crucial due to its direct linkage with riding comfort
Published: 5 June 2023
and user costs [9]. It also has a substantial impact on vehicle dynamics. Pavement roads of
higher roughness increase users’ operational costs, fuel consumption, and tire wear, and
reduce vehicle durability [9–11]. Lui and Al-Qadi [12] found that increasing pavement
Copyright: © 2023 by the authors.
roughness corresponds to an increase in the impact of dynamic loading on fuel consumption.
Licensee MDPI, Basel, Switzerland.
Robbins and Tran [13] reported that lowering the International Roughness Index (IRI) from
This article is an open access article 4 m/km to 3 m/km reduces vehicle maintenance costs by about 10%. Additionally, reducing
distributed under the terms and 1 m/km in IRI can save USD 340 million in tire wear costs for passenger cars [13]. Moreover,
conditions of the Creative Commons the impact of roughness on accident rates was remarked by different studies [14–16]. Thus,
Attribution (CC BY) license (https:// it is usually used to capture commuters’ experience on roads.
creativecommons.org/licenses/by/ Pavement roughness is typically categorized into three scales based on functional con-
4.0/). siderations, such as safety and riding quality. The three scales of pavement roughness are

Remote Sens. 2023, 15, 2941. https://doi.org/10.3390/rs15112941 https://www.mdpi.com/journal/remotesensing

 passenger cars [13]. Moreover, the impact of roughness on accident rates was remarked
by different studies [14–16]. Thus, it is usually used to capture commuters’ experience on
roads.
Pavement roughness is typically categorized into three scales based on functional
Remote Sens. 2023, 15, 2941 considerations, such as safety and riding quality. The three scales of pavement roughness 2 of 44
are roughness, macrotexture, and microtextured. This paper focuses on the pavement
roughness of the largest scale, which is defined as the deviation of pavement surface from
the true planner surface. Roughness refers to surface irregularities with characteristics
roughness, macrotexture, and microtextured. This paper focuses on the pavement roughness
wavelength ranging between 0.1 and 100 m and amplitudes ranging from 1 to 100 mm
of the largest scale, which is defined as the deviation of pavement surface from the true
[17,18]. It is typically measured in the wheel path area [9]. It primarily impacts vehicle
planner surface. Roughness refers to surface irregularities with characteristics wavelength
dynamics, vehicle 0.1
ranging between wear,
androad100 m hold,
andand riding comfort
amplitudes ranging[9,17,18].
from 1 to In100this
mm study, theItterms
[17,18]. is typi-
“pavement roughness” and “roughness” are used interchangeably
cally measured in the wheel path area [9]. It primarily impacts vehicle dynamics, vehicle to refer to the highest-
wear,
scale
roadpavement
hold, and roughness.
riding comfort [9,17,18]. In this study, the terms “pavement roughness” and
Traditionally,
“roughness” rideinterchangeably
are used quality was measured to refer tobythe
a group of raterspavement
highest-scale travelingroughness.
the road and
subjectively giving verbal
Traditionally, or quantitative
ride quality rates, such
was measured by a as in the
group ofPavement Serviceability
raters traveling the roadRat- and
ing (PSR) [19]. Over time, various measures were used to evaluate
subjectively giving verbal or quantitative rates, such as in the Pavement Serviceability roughness, including
IRI, Profile
Rating Index
(PSR) (PI),
[19]. Over Ride Quality
time, Index
various (RQI), were
measures and Half-car Roughness
used to evaluate Index (HRI)
roughness, [20].
including
Power
IRI, Profile Index (PI), Ride Quality Index (RQI), and Half-car Roughness Index (HRI)and
spectral density (PSD) is also widely used for pavement roughness evaluation [20].
constitutes the basis
Power spectral for ISO
density (PSD)road profile
is also classification
widely used for [21].
pavementIn theroughness
1980s, IRI and HRI were
evaluation and
the most used
constitutes themeasures
basis for ISO of roughness
road profile in classification
the USA [22]. However,
[21]. the Federal
In the 1980s, IRI andHighway
HRI were
Agency’s
the most(FHWA) efforts to
used measures of standardize
roughness inthe theroughness
USA [22].measurement
However, thebroughtFederalIRI to be
Highway
universally accepted in the USA and elsewhere [23]. IRI is a
Agency’s (FHWA) efforts to standardize the roughness measurement brought IRI to be statistical method to report
the roughness
universally of pavement.
accepted in the ItUSAis mathematically
and elsewhere [23].computed
IRI is afrom a single
statistical longitudinal
method to reportpro-the
file by applying
roughness a reference
of pavement. It isquarter car simulation
mathematically computedshown in Figure
from a single1.longitudinal
IRI was initiated
profile
based on extensive
by applying research
a reference conducted
quarter by the World
car simulation shown Bank. It is reported
in Figure in units
1. IRI was of inches
initiated based
per
on mile or meters
extensive per kilometer
research conducted [23].
by the World Bank. It is reported in units of inches per
mile or meters per kilometer [23].

Figure
Figure1.1.Quarter
Quartercar
carmodel
model[24].
[24].

Roughnessdata
Roughness datacollection
collection techniques
techniques have
have experienced
experiencedthe thefastest
fastestpace
paceofofautomation
automa-
andand
tion technology
technology maturity
maturity[25–27].
[25–27].However,
However,developing roughnessmeasurement
developing new roughness measurement
techniquescontinues
techniques continuesto to capture
capture the attention of of many
manyresearchers
researchersdue duetotoa avariety
variety ofof
factors.
fac-
One crucial driver was the need to create efficient, easy-to-integrate, and highly
tors. One crucial driver was the need to create efficient, easy-to-integrate, and highly au- automated
data collection
tomated means.means.
data collection Some Some
researchers have have
researchers focused on improving
focused on improving the the
accuracy and
accuracy
precision
and of the
precision collected
of the data.
collected Additionally,
data. Additionally, several research
several researchefforts have
efforts haveprioritized
prioritizedthe
development
the development of of
cost-effective and
cost-effective low-cost
and low-costtechniques.
techniques.
Moreover,the
Moreover, thegrowing
growingdemand
demandfor forroughness
roughnessmeasurement
measurementby bynew
newindustries
industrieshas has
brought forward unique needs. Highway agencies are no longer the sole party interestedin
brought forward unique needs. Highway agencies are no longer the sole party interested
collecting and using roughness data. Measuring and using roughness data are increasingly
becoming an interest of the automotive industry and normal road users. Autonomous
vehicles are expected to have the capacity to measure pavement roughness to ensure safety
and improve riding comfort. Additionally, roughness data are increasingly becoming
important for developing more efficient navigation systems. Thus, new techniques are
required to fulfill the needs of more diverse industries.
Remote Sens. 2023, 15, 2941 3 of 44

Several publications have reviewed the research efforts related to various aspects of
pavement roughness. Nguyen et al. [28] reviewed the use of response-type methods for
evaluating pavement surface anomalies. The review covered 130 studies on the use of
response-type methods for surveying pavement defects published between 2006 and 2019.
The authors reported a significant focus on using smartphones, connected vehicles, and
machine learning for pavement profile and roughness estimation. Robbins and Tran [10]
and Zaabar and Chatti [11] reviewed the impact of pavement roughness on vehicles’
operating costs. Pavement roughness was concluded to have a measurable increase in fuel
consumption and maintenance costs [10,11]. Hettiarachchi et al. [29] reviewed roughness
indices used in different countries around the world and indicated the prevalence of using
IRI. In a recent study, Yu et al. [30] examined the use of smartphones in measuring pavement
roughness and other surface anomalies.
However, there is a lack of literature devoted to understanding the evolution of
pavement measurement techniques in industry practice and research. Additionally, it is
vital to recognize the current trends and needs for pavement roughness measurement to
better direct future research. Thus, this study tracked the changes over the past five decades
to analyze the evolution of pavement roughness measurement techniques in academia
and industry practice. The evolution of pavement roughness technologies was reviewed
primarily in North America, using the TRID library, as a resemblance to the industry
advancements worldwide. The state of practice was also analyzed by reviewing dozens of
pieces of equipment in eleven countries in North America, Europe, and Australia. Research
trends were analyzed using the Scopus database.

2. Research Methodology
The methodology of the current study is outlined in Figure 2. The publications used
for conducting the analysis were identified in two subsequent stages. In the first stage,
Scopus and TRID Library databases were searched to identify relevant publications. The
Scopus database is selected to retrieve relevant research articles. Scopus is recognized to be
the largest and most comprehensive literature database [31,32]. The search in the TRID was
restricted to the TRB (Transportation Research Board) Online Bookstore. The TRB Library
was founded in 1946 as the primary archive of the Transportation Research Board, Highway
Research Board, Strategic Highway Research Program, and the Marine Board. The library
has a vast collection of research articles as well as technical reports and monographs [33].
The Scopus database was primarily used to retrieve relevant academic research articles,
whereas the TRID library was used to retrieve syntheses describing industry practices.
As shown in Figure 2, the search query of “(pavement* OR asphalt OR road* OR
street*) AND (roughness OR IRI OR profile OR profiler* OR profilometer*)” was utilized to
retrieve the relevant literature in the Scopus database. The asterisk “*” was used to account
for the different variants of the keywords (e.g., the plural form). The search fields were
limited to article titles and keywords to obtain the most relevant publications. To obtain
syntheses describing the industry practices, the search was conducted using the TRID [34].
The keyword “pavement” was used as a search query to retrieve relevant publications. All
publications with “pavement” in their titles were retrieved. As a result, 3236 and 192 items
were retrieved from Scopus and TRID, respectively.
In the second stage, the obtained research results were refined. In the Scopus database,
the refining process included three steps. Firstly, search results were filtered to only include
original journal articles focused on engineering and written in English. This refinement
resulted in eliminating more than 60% of the original search results. Then, the articles
were investigated in two steps to only include articles focused on developing, improving,
or testing new techniques for pavement roughness measurement. A total of 126 articles
were identified, and four more were included during the analysis, bringing the number of
considered articles to 130.
Remote
Remote Sens.
Sens. 2023,
2023, 15,
15, x2941
FOR PEER REVIEW 5 4of
of 45
44

Figure
Figure 2. Research
Research methodology.
methodology.

In the TRID Analysis

3. Scientometric Library (TRB Online Bookstore), the search was limited to the Highway
(NCHRP) category. The National Cooperative Highway Research Program (NCHRP) is
3.1. Syntheses of Practice
an objective national highway research program launched in 1962 in the United States of
This study
America. makes
It focuses onuse of seven NCHRP
conducting researchsyntheses of practice
to help improve the published
way nationalbetween 1981
highways
and 2022, as presented in Table 1. These provided valuable insights into the
are designed, built, operated, and maintained. Syntheses of Highway Practice are used to shifts in pave-
ment roughness
complement the data collection
original over
research the previous
efforts five decades
by documenting in North
the state America.
of practice Infor-
regarding
mation related to the collection of roughness data was extracted to provide
a given topic. Synthesis reports are often based on literature reviews and comprehensive a comprehen-
sive picture
surveys of of the evolution
activities of pavement
and initiatives. roughness
NCHRP hasmeasurement
published more in the past400
than fewsynthesis
decades.
The extracted
reports information
covering covers the
various aspects popularity[35].
of highways of pavement roughness
Limiting the search todata
thecollection
documents as
well as the used techniques, as presented in Table 1.
categorized as Highway (NCHRP) resulted in the elimination of over 60% of the original
search output. After skimming the remaining publications’ content, only six items were
Table 1. Summary
selected of the
to conduct theanalyzed
analysis.NCHRP Synthesis of Highway Practice.
In the final stage, the obtained results were
Roughness Dataanalyzed to identify trends in both industry
Collection
Synthesis and academic research
Year over the previous half-century. Temporal Roughness Measure-
analysis revealed that
Spatial Temporal Reporting
retrieved publications covered a period of approximately fifty years. ment
TheTechniques
trends were
Coverage Coverage Unit
analyzed over four time periods spanning between 1971 and 2022. The periods were
NCHRP 76: Collection and Use of Pavement
segmented based on1981 the availability
√ of syntheses√ of practice. - A bibliometric analysis
√ was
Condition Data first
[19] conducted for Scopus literature using the Shiny R package [36]. Then, both the Scopus
NCHRP 126: Equipment for andObtaining Pave-
TRID Library-based
1986 publications
√ were analyzed
- to identify
- the trends in√ pavement
ment Condition and Traffic Loading Data [25]
roughness measurement techniques over the last half-century. In this study, pavement
NCHRP 203: Current Practices in Determining
roughness techniques were analyzed in terms of their directness of measuring pavement
1994 √ √ √ √
Pavement Condition roughness,
[27] accuracy, cost-effectiveness and level of automation.
NCHRP 334: Automated Pavement Distress
2004 √ √ √ √
Collection Techniques [23]
Remote Sens. 2023, 15, 2941 5 of 44

3. Scientometric Analysis
3.1. Syntheses of Practice
This study makes use of seven NCHRP syntheses of practice published between 1981
and 2022, as presented in Table 1. These provided valuable insights into the shifts in
pavement roughness data collection over the previous five decades in North America.
Information related to the collection of roughness data was extracted to provide a com-
prehensive picture of the evolution of pavement roughness measurement in the past few
decades. The extracted information covers the popularity of pavement roughness data
collection as well as the used techniques, as presented in Table 1.
Table 1. Summary of the analyzed NCHRP Synthesis of Highway Practice.

Roughness Data Collection Roughness

Synthesis Year Spatial Temporal Reporting Measurement
Coverage Coverage Unit Techniques
NCHRP 76: Collection and Use of √ √ √
1981 -
Pavement Condition Data [19]
NCHRP 126: Equipment for √ √
Obtaining Pavement Condition and 1986 - -
Traffic Loading Data [25]
NCHRP 203: Current Practices in √ √ √ √
Determining Pavement 1994
Condition [27]
NCHRP 334: Automated Pavement √ √ √ √
2004
Distress Collection Techniques [23]
NCHRP 401: Quality Management √ √ √ √
of Pavement Condition Data 2009
Collection [37]
NCHRP 589: Automated Data
Collection and Quality √ √ √ √
2017
Management for Pavement
Condition Reporting [26]
NCHRP 589: Automated Data
Collection and Quality √ √
2022 * *
Management for Pavement
Condition Reporting [2]
√
Information was included; - Information was missing; * All agencies in the USA were assumed to be entitled to
PM2 [38] reporting requirements.

The syntheses are based on extensive surveys designed to capture the practices in
pavement data collection and management in the USA and Canada. The number of agencies
surveyed in the seven utilized syntheses is shown in Figure 3.
Considering the available information regarding the industry practices, four study
periods were segmented to analyze trends in the industry and academic research. The first
study period extends to 1986, during which two syntheses were identified. The second
study period extends from 1987 to 1994. Information regarding industry practice in this
period is analyzed using the third synthesis. The third study period spans from 1995 to
2009. Two syntheses are available to provide insights regarding industry practices in this
period, published in 2004 and 2009. The last study period also includes two syntheses and
covers a span of 13 years, from 2010 to 2022.
 √ Information was included; - Information was missing; * All agencies in the USA were as
be entitled to PM2 [38] reporting requirements.

The syntheses are based on extensive surveys designed to capture the pra
Remote Sens. 2023, 15, 2941 pavement data collection and management in the USA and Canada. The number
6 of 44

cies surveyed in the seven utilized syntheses is shown in Figure 3.

Figure
Figure 3. Number
3. Number of surveyed
of surveyed agencies
agencies in seven
in the used the used seven syntheses.
syntheses.

3.2. Journal Articles

Considering the available information regarding the industry practices, fo
This study makes use of 130 original articles. The retrieved articles were published
over a timewere
periods segmented
period toaanalyze
of about half trends in the
century (1971–2022). Theindustry and academic
annual growth research.
is about 6.2%,
study period
whereas extends
the documents’ to 1986,
average during
age is which
about 12 years.two syntheses
Figure 4 presentswere identified.
the number of The
articles published in the four segmented study periods. The vast majority
study period extends from 1987 to 1994. Information regarding industry practic
Remote Sens. 2023, 15, x FOR PEER REVIEW (77%) of the
articles were published in the fourth period (2010–2022), with a yearly production rate of
period is analyzed using the third synthesis. The third study period spans from
about eight articles. The lowest number of articles was retrieved from the third period
2009. Twowith
(2004–2009), syntheses
an annualare available
research to of
output provide insights regarding industry practice
0.4 articles.
period, published in 2004 and 2009. The last study period also includes two synth
covers a span of 13 years, from 2010 to 2022.

3.2. Journal Articles

This study makes use of 130 original articles. The retrieved articles were p
over a time period of about half a century (1971–2022). The annual growth is abo
whereas the documents’ average age is about 12 years. Figure 4 presents the nu
articles published in the four segmented study periods. The vast majority (77%
articles were published in the fourth period (2010–2022), with a yearly productio
about eight articles. The lowest number of articles was retrieved from the third
(2004–2009), with an annual research output of 0.4 articles.

Figure 4. Periodic and cumulative number of publications.

Figure 4. Periodic and cumulative number of publications.
3.2.1. Most Relevant Sources
The
3.2.1. retrieved
Most articles Sources
Relevant were analyzed using the Shiny R package [36] to identify the
most relevant sources. The analysis revealed 67 different sources; most of them (76%)
Thea retrieved
circulated articles
single article. The top were analyzed
nine sources using
of at least threethe Shiny
articles R package
are presented in [36] t
most relevant sources. The analysis revealed 67 different sources; most of th
Table 2. As presented in Table 2, the “Transportation Research Record” and “International
Journal of Pavement Engineering” circulated the highest number of articles. In total, they
culated a single article. The top nine sources of at least three articles are pres
circulated about a quarter of the retrieved articles.
2. As presented in Table 2, the “Transportation Research Record” and “Intern
of Pavement Engineering” circulated the highest number of articles. In total, t
about a quarter of the retrieved articles.
The core collection of the most relevant journals according to the Sc
Remote Sens. 2023, 15, 2941 7 of 44

Table 2. Most relevant sources.

Source No. Articles Core Collection

Transportation Research Record 17 Civil Engineering; Transportation Science & Technology
Civil Engineering; Materials Science, Characterization,
International Journal of Pavement Engineering 15
Testing; Construction & Building Technology
Instruments & Instrumentation; Electrical & Electronic
Sensors 8
Engineering
Journal of Transportation Engineering Part B: Pavements 6 Civil Engineering; Transportation Science & Technology
Civil Engineering; Computer Science, Interdisciplinary
Journal of Computing in Civil Engineering 5
Applications
Journal of Infrastructure Systems 4 Civil Engineering
Vehicle System Dynamics 4 Mechanical Engineering
Automation in Construction 3 Civil Engineering; Construction & Building Technology
Journal of Testing and Evaluation 3 Materials Science, Characterization & Testing

The core collection of the most relevant journals according to the Science Citation
Index Expanded (SCIE) is also presented in Table 2. Analysis of the core collection subject
categories of the most relevant sources based on the SCIE revealed some diversity. Most of
the relevant sources are specialized in civil and engineering, transportation science, and
technology. However, the core collection of top sources contains other subject categories
such as construction and building technology, sensing, instrumentation, computer science,
and mechanical engineering.

3.2.2. Top Relevant Authors’ Affiliations

The authors’ affiliations are analyzed to evaluate the spread of relevant research. In
total, 114 affiliations were identified for the 130 analyzed articles. Table 3 presents the eight
top relevant authors’ affiliations. About 60% of the identified affiliations are associated
with one article, whereas about 12% are associated with at least five articles. Top authors’
affiliations constitute universities in the USA, China, and Ireland. North Dakota State
University and Tongji University are the most affiliated institutions among the relevant
authors. Table 3 indicates that all of the articles affiliated with the Pennsylvania State
University (USA) were published in the second period. In contrast, all the articles affiliated
with Chinese universities have been published in the past few years.
Table 3. Top relevant authors’ affiliations.

Cum. No. of Articles over the Four Studied Periods

Affiliations Country
I II III IV
North Dakota State University USA 0 0 0 16
Tongji University China 0 0 0 15
Pennsylvania State University USA 1 10 10 10
University College Dublin Ireland 0 0 4 8
Jiangsu University China 0 0 0 7
Beihang University China 0 0 0 7
Civil Aviation University of China China 0 0 0 7
Microwaves and Radar Institute Germany 0 0 0 7
 Pennsylvania State University USA 1 10 10 10
University College Dublin Ireland 0 0 4 8
Jiangsu University China 0 0 0 7
Beihang University China 0 0 0 7
Civil2023,
Remote Sens. Aviation
15, 2941University of China China 0 0 0 7 8 of 44
Microwaves and Radar Institute Germany 0 0 0 7

3.2.3. Corresponding
3.2.3. CorrespondingAuthor’s
Author’sCountries
Countries
Figure 5 shows the corresponding
Figure 5 shows the corresponding author’s country.
author’s Additionally,
country. it shows
Additionally, the num-
it shows the num-
ber of single-country publications (SCP), which include only intra-country collaboration,
ber of single-country publications (SCP), which include only intra-country collaboration,
and the multiple-country publications (MCP), which include inter-country collaboration.
and the multiple-country publications (MCP), which include inter-country collaboration.
Figure 5 indicates that most of the articles corresponded to authors from the USA (37) and
Figure 5 indicates that most of the articles corresponded to authors from the USA (37) and
China (22). Moreover, it is remarkable that most of the articles are SCP, indicating limited
China (22). Moreover, it is remarkable that most of the articles are SCP, indicating limited
inter-country collaboration. In fact, just about 16% of the articles involve intra-country
inter-country collaboration. In fact, just about 16% of the articles involve intra-country
collaboration. The collaborations were mainly found between researchers from the USA
collaboration. The collaborations were mainly found between researchers from the USA
and China on the one hand, and other countries, including Poland, the United Kingdom,
and China on the one hand, and other countries, including Poland, the United Kingdom,
Iran, and Jordan, on the other.
Iran, and Jordan, on the other.

Figure Number
5.Number
Figure 5. of of publications
publications for top
for the the affiliated
top affiliated countries
countries considering
considering the corresponding
the corresponding au-
thor’s country.
author’s country.

3.2.4. Top Cited

3.2.4. Top CitedPapers
Papers
Table 44 presents
Table presentsdetails
detailsofofthe
themost
most cited
cited articles
articles among
among the the retrieved
retrieved documents.
documents.
“The use of vehicle acceleration measurements to estimate road roughness”
“The use of vehicle acceleration measurements to estimate road roughness” [39] is by [39]
faris by
far the most cited article, with 200 citations. The article explored using vehicles’
the most cited article, with 200 citations. The article explored using vehicles’ built-in sen- built-in
sensors for pavement
sors for pavement roughness
roughness estimation.
estimation. This indicates
This indicates the in
the surge surge in research
research focusing
focusing on
on
equipping vehicles, particularly highly automated and autonomous vehicles, with pave-with
equipping vehicles, particularly highly automated and autonomous vehicles,
pavement roughness
ment roughness measurement
measurement capabilities.
capabilities. Othercited
Other highly highly cited
papers papersonfocused
focused using on
using accelerometers, smartphones, and vehicles’ built-in sensors to build systems for
pavement roughness estimation. It is worth noting that five of the seven most cited articles
were published in journals with an SCIE core collection coverage of mechanics, electronics,
sensing, automation, and technology rather than transportation and civil engineering.
This indicates a growing interest in measuring pavement roughness by researchers from
disciplines outside the classical axis of civil engineering and transportation.
Table 4. Top cited articles.

Article Year Technology Experiment Type Source Total Citations

[39] 2008 Vehicles’ built-in sensors Simulation Vehicle System Dynamics 200
Transportation Modeling and
[40] 2014 Accelerometers-based system Field 74
Management
[41] 2018 Smartphone-based system Field Sensors 68
In-pavement strain sensors
[42] 2020 Field Journal of Lightwave Technology 67
(optical fiber)
Simulation, accelerometers-based
[43] 2010 Simulation Journal of Sound and Vibration 62
system
Simulation using IEEE Transactions on Control
[44] 2015 Accelerometers-based system 57
testbed Systems Technology
Transportation Research Part C:
[45] 2021 Vehicles’ built-in sensors Field and simulation 57
Emerging Technologies
Remote Sens. 2023, 15, 2941 9 of 44

3.2.5. Keywords Analysis

Table 5 presents the fifteen author keywords with the most occurrence and total link
strength. Additionally, a word cloud is generated to visualize the occurrence of author
keywords, as shown in Figure 6. To derive a better characterization of the retrieved articles,
synonym keywords were combined. For example, “smartphone” and “smartphones” were
grouped as one keyword. Similarly, “roughness”, “road roughness”, “street roughness”,
and “pavement roughness” were also considered as a single keyword. A complete list
of the synonym keywords is presented in the Appendix A. As presented in Table 5 and
illustrated in Figure 6, “pavement roughness” and “international roughness index” are by
far the most used keywords by authors. The first keyword makes a general indication of
the research field. The second keyword indicates the popularity of using IRI to measure
pavement roughness. The keywords “Smartphones”, “accelerometers”, “connected vehi-
cles”, and “crowdsourcing” appear as the fourth, fifth, tenth, and eleventh most frequently
occurring keywords. They represent a contemporary research trend in pavement roughness
evaluation. Other keywords such as “Pavement management”, “ride quality”, “pavement
profiles”, “pavement condition assessment”, and “Road Roughness Estimation” constitute
the main objectives of measuring pavement roughness. “Power spectrum density” is widely
used by researchers focused on studying pavement roughness in combination with car ride
quality and vehicle suspension. The presence of “Machine Learning” and “Artificial Neural
Network” (ANN) among the most frequently used keywords implies the importance of
using machine learning for pavement roughness evaluation.
Table 5. Most occurred author keywords.

Keyword Occurrence Total Link Strength

Pavement Roughness 46 209
International Roughness Index 35 158
Pavement Management 12 56
Remote Sens. 2023, 15, x FOR PEER REVIEW Smartphones 10 10 of
45 45
Accelerometers 9 45
Artificial Neural Network 8 42
Artificial Neural Network 8 42
Power Spectral Density 8 40
Power Spectral Density 8 40
Ride Quality 8 33
Ride Quality 8 33
Connected
Connected Vehicles
Vehicles 7 7 37 37
Pavement Profiles
Pavement Profiles 7 7 28 28
Crowdsourcing
Crowdsourcing 6 6 25 25
Pavement
PavementCondition
ConditionAssessment
Assessment 6 6 29 29
Machine Learning
Machine Learning 5 5 30 30
Road Roughness Estimation 5 22
Road Roughness Estimation 5 22
Vehicle Responses 5 18
Vehicle Responses 5 18

Figure Wordcloud
Figure 6.6.Word cloud
of of
thethe most
most frequently
frequently occurring
occurring authorauthor keywords.
keywords.

4. Trends of Pavement Roughness Measurement Equipment

The collection of roughness data has experienced different gradual shifts in the last
few decades. Figure 7 attributes the data collection practices over the last five decades.
Remote Sens. 2023, 15, 2941 10 of 44
Figure 6. Word cloud of the most frequently occurring author keywords.

4. Trends of Pavement Roughness Measurement Equipment

4. Trends of Pavement Roughness Measurement Equipment
The collection of roughness data has experienced different gradual shifts in the last
The collection
few decades. Figureof7 roughness datadata
attributes the has collection
experienced different
practices gradual
over the lastshifts
fivein the last
decades.
few decades. Figure 7 attributes the data collection practices over the last
Roughness has been widely evaluated qualitatively since at least the end of the 1970s. Thefive decades.
Roughnessofhas
collection been widely
roughness dataevaluated
has beenqualitatively
continuouslysince at least
popular the end
among of the 1970s.
highway The
agencies.
collection of roughness data has been continuously popular among highway
However, the collection of roughness data at the network level was less practiced during agencies.
However,
the the collection
1980s. Due of roughness
to the availability data at thethat
of equipment network level
enables thewas less practiced
collection during
of roughness
the 1980s. Due to the availability of equipment that enables the collection of roughness data
data at the traffic level, sampling was rarely practiced. Additionally, the frequency of
at the traffic level, sampling was rarely practiced. Additionally, the frequency of roughness
roughness data collection increased over the years to peak on an annual and biannual
data collection increased over the years to peak on an annual and biannual basis. The use
basis. The use of IRI for reporting roughness has gradually been mainstreamed and be-
of IRI for reporting roughness has gradually been mainstreamed and became prevalent
came prevalent during the third and fourth periods.
during the third and fourth periods.

Remote Sens. 2023, 15, x FOR PEER REVIEW 11 of 45

Figure7.
Figure Popularityof
7. Popularity ofroughness
roughnessdata
data collection.
collection.
4.1. Roughness Data Collection
4.1. Roughness Data Collection
The popularity of different types of roughness measurement equipment in North
The over
America popularity
the lastoffivedifferent
decades types of roughness
is summarized measurement
in Figure equipment
8. The use in North
of subjective eval-
America
uation was over
rarethe lastinfive
even the decades is summarized
1980s. Multiple in Figure
agencies used 8. The use
profilographs andofcontact
subjective
pro-
evaluation was rare even in the 1980s. Multiple agencies used profilographs
filers in the first period. Some types of contact profilers continue throughout the second and contact
profilers in thebeing
period before first period.
phasedSome types of
out during thecontact profilersResponse-type
third period. continue throughout
methods thewere
second
the
period before being phased out during the third period. Response-type
primary equipment for measuring roughness during the first and second periods. How- methods were the
primary
ever, theyequipment for measuring
were gradually replacedroughness during the
by non-contact first and
profilers. secondthe
During periods. However,
third study pe-
they were gradually replaced by non-contact profilers. During
riod, response-type methods were entirely replaced by the non-contact profilers. the third study period,
Non-
response-type
contact profilers methods wereand
of infrared entirely replaced
ultrasonic by the
sensors non-contact
were profilers.
quite popular duringNon-contact
the second
profilers of infrared
period. However, andwere
they ultrasonic sensors
gradually werewith
replaced quiteequipment
popular during the second
that comprises period.
laser sen-
However, they were gradually replaced with equipment that comprises
sors. The use of 3D imagining systems started garnering attention about two decades ago. laser sensors. The
use of 3D imagining systems started garnering attention about two decades ago. Today,
Today, many agencies utilize such systems for pavement roughness measurements.
many agencies utilize such systems for pavement roughness measurements.

Figure 8. Popularity
Figure 8. Popularity of
of roughness
roughness data
data collection
collection technologies
technologies in
in industry
industry practice.
practice.

Period I: Until 1986

Before the 1950s, roughness data were seldom collected at the network level [23].
However, the collection of roughness data has gradually become more common in recent
Remote Sens. 2023, 15, 2941 11 of 44

Period I: Until 1986

Before the 1950s, roughness data were seldom collected at the network level [23].
However, the collection of roughness data has gradually become more common in re-
cent decades. In the early 1980s, roughness was among the data collected as part of
network-wide surveys conducted by some agencies in North America. Although the 1981
synthesis [19] was limited to surveying nine agencies, the findings give indications of the
increasing trend of wide-scale collection of condition data, including roughness. In the
1981 synthesis [19], eight of the nine surveyed agencies (89%) reported collecting quanti-
tative roughness data. Differently, one agency (Ontario) reported subjective ride quality
measures. The majority of the surveyed agencies (67%) collected roughness data for 100%
of the network every one or two years. Only USA Air Force did not adopt programmed
roughness data collection campaigns [19].
The collection of pavement condition data experienced a remarkable boost in the first
half of the 1980s, as reported in the 1986 synthesis [25]. The 1986 synthesis captured a
significant surge in the number of agencies collecting condition data on a large scale [25].
Roughness data were among the most frequently gathered by agencies. Out of the 44
responses collected from highway agencies in the USA and Canada, 40 agencies confirmed
the collection of roughness data [25]. However, only 27 agencies (61%) collected roughness
data for pavement management purposes at the network level. Seven other agencies
(16%) reported occasional use of such data for pavement management at the network level.
Some agencies used the collected data for other multiple functions, including research,
planning, and pavement management at the project level [25]. However, no information
was provided on the frequency and coverage of the roughness data collection surveys.
Period II: 1987–1994
The collection of roughness data experienced a dramatic boost in the late 1980s and
early 1990s. This remarkable surge can be largely attributed to the 1987 update of the
HPMS (Highway Performance Monitoring System) field manual [46]. The 1987 update of
the HPMS field manual required a collection of pavement roughness data at least every
two years [46]. The effect of this landmark requirement was vivid in the survey results
published in the 1994 synthesis [27]. Fifty-nine agencies (98%) reported collecting roughness
data at the network level for at least part of their network. Tennessee in the USA was the
only agency that did not report conducting roughness surveys. Thirty-eight (64%) agencies
reported collecting roughness data on the entire network system. Anther ten agencies
stated collecting roughness data on the part of their networks (e.g., interstate, primary
roads, secondary roads).
Additionally, the use of sampling for collecting roughness data was minimal, with only
one agency reporting adopting some sort of sampling. The frequency of collecting pavement
roughness data was also notable. Twenty-two (37%) and thirteen (22%) agencies were collecting
roughness data for the entire system every one or two years, respectively [27]. A more
significant portion of surveyed agencies (68%) collect roughness data on the interstate network.
Furthermore, all agencies were required to report roughness using IRI for the HPMS [46].
The requirement of submitting the roughness data in the standardized form of IRI had a
great benefit that allowed for the comparison of roughness data between different highway
agencies. Forty-one (68%) agencies measured or converted the roughness measurement to
IRI. Other units include ride number (RN), riding comfort index (RCI), root-mean-square
vertical acceleration (RMSVA), and road roughness index (RRI) [22,47].
Period III: 1995–2009
The popularity of roughness data collection reached a new peak by the 2004 synthe-
sis [23] when all surveyed agencies started collecting roughness data from at least part
of their network system. Twenty-six (46%) agencies collected roughness data annually,
whereas twenty-four (43%) agencies collected roughness data every two to three years.
Some agencies reported higher collection frequency on interstate networks compared to
other parts of the network. For example, Oregon, Ohio, and Alabama reported collecting
Remote Sens. 2023, 15, 2941 12 of 44

roughness data on interstate networks annually, whereas the data was collected for other
parts of the network every two years. No sampling was employed for collecting pavement
roughness data. Instead, the total length of the surveyed evaluation road lane was surveyed.
Fifty-two (93%) agencies reported using automated tools. Another remarkable finding was
the adoption of IRI as a standardized form of the roughness data by all of the agencies that
responded [23]. The HPMS necessitated reporting pavement roughness using IRI for the
entire National Highway System road network every two years [23].
Similarly, the 2009 synthesis [37] reported that about 95% of the surveyed agencies
in the USA and Canada collected roughness data at the network level. Thirty-two (58%)
agencies were found to collect roughness data for the highway system every year. Another
seventeen (31%) agencies reported collecting roughness data on the highway system every
two to three years. Remarkably, the collected distress data including roughness, have
increasingly utilized to calibrate mechanistic–empirical pavement design using pavement
models based on the mechanics of material [48]. The frequency of collecting roughness
data kept rising primarily due to the requirement of HPMS Reassessment 2010+ [49]. The
HPMS Reassessment 2010+ required submitting roughness data for the National Highway
System each year [37]. The synthesis reiterates the widespread use of IRI for measuring
pavement roughness [37].
Period IV: 2010–2022
The 2019 and 2022 syntheses [2,26] reported that virtually all agencies were collecting
roughness data for the entire network on a yearly basis. Furthermore, IRI continued
to be used as a standardized measure of pavement roughness in North America [2,26].
The increase in the popularity of pavement roughness data collection has primarily been
fueled by Federal regulations [23 CFR 490 National Performance Management Measures
(PM2)] [38]. The PM2 requires state highway agencies in the USA to collect and report IRI,
rutting, faulting, and crack percentage for the Interstate Highway System and the Non-
Interstate National Highway System [38]. The PM2 requires reporting pavement condition
data every one and two years for the interstate and non-interstate national highway system
(NHS) systems, respectively.

4.2. Roughness Measurement Technologies

4.2.1. Popular Equipment
Various techniques have been used to evaluate pavement roughness over time. In
the early days, simple straightedges were used as the sole means to evaluate pavement
roughness. Over time, enormous efforts and extensive research were focused on developing
more sophisticated equipment to measure roughness. In the first half of the twentieth
century, different pieces of equipment with varying levels of sophistication were designed.
Initially, the developed equipment was primarily mechanical with elaborated multi-wheel
support systems, usually referred to as profilographs. With time, advanced technologies
were employed to develop devices that incorporate mechanical systems, electrical circuits,
electronics, accelerometers, infrared, ultrasonics, and lasers [50].
Pavement roughness equipment can be categorized based on different considerations.
However, roughness measurement equipment can broadly be classified into two main
groups based on their application principle. The first category constitutes equipment that
is used for direct measurement of pavement profile. As shown in Figure 9, such equipment
type includes profilographs, laser profilers, and 3D imaging systems (e.g., laser-based
imaging systems and depth cameras). The longitudinal profile is evaluated using the col-
lected data to ultimately calculate the roughness index, usually expressed as an IRI value.
The second category includes equipment that indirectly measures the pavement profile.
Alternatively, they measure vehicle response to pavement profile irregularities. Figure 10
presents some acceleration-based equipment: conventional response-type equipment (e.g.,
Mays meter), smartphones, mounted accelerometers, and vehicles’ built-in sensors. The
collected acceleration signals are double-integrated to obtain the displacement that is used
for evaluating the longitudinal profile [51]. The calculated roughness index is correlated
RemoteRemote Sens. 2023,
Sens. 2023, 15, x 15,
FOR 2941
PEER REVIEW 13 of 44
14 of 45

Remote Sens. 2023, 15, x FOR PEER REVIEW 14 of 45

with the corresponding values measured using high-accuracy equipment such as laser pro-
filers [52]. In this study, the latter type of equipment is tagged as response-type equipment.

Figure 9. Roughness
Figure 9. Roughnessevaluation
evaluation using directmeasurement
using direct measurement equipment
equipment typetype
[53].[53].
Figure 9. Roughness evaluation using direct measurement equipment type [53].

Figure 10. Roughness evaluation using response-type methods (accelerometer-based) [54].

World Bank classification is another widely used categorization of pavement rough-

Figure
ness10.
Figure Roughness
10. Roughnessevaluation
measurement equipment using
evaluation using response-type
[52]. response-type
The methods
differentmethods
types (accelerometer-based)
of(accelerometer-based)
equipment [54]. [54].
were classified con-
sidering their precision measurement of IRI. Class I constitutes precision profilers with
theWorld
highestBank classification
roughness is another
measurement widely
accuracy. used categorization
However, ofare
Class II profilers pavement
of a lowerrough-
ness
accuracy than those of Class I. Class III includes mechanical or electronic equipment to con-
measurement equipment [52]. The different types of equipment were classified
sidering their precision measurement of IRI. Class I constitutes precision profilers with
Remote Sens. 2023, 15, 2941 14 of 44

World Bank classification is another widely used categorization of pavement roughness

measurement equipment [52]. The different types of equipment were classified considering
their precision measurement of IRI. Class I constitutes precision profilers with the highest
roughness measurement accuracy. However, Class II profilers are of a lower accuracy than
those of Class I. Class III includes mechanical or electronic equipment to evaluate the pavement
roughness indirectly. These devices are calibrated by correlating their measures with standard-
ized roughness values. Class IV includes subjective evaluation of pavement roughness via
visual inspection or by driving over the pavement section. Class IV represents a qualitative
assessment based on guidance and personal experience. A brief description of some of the
previously and currently popular roughness techniques in North America is presented.
Profilographs
Profilographs consist of a beam or frame carried by a system of wheels in addition
to a profile wheel to measure pavement roughness. The supported wheels, which vary in
number according to the profilograph type, carried the beam or the frame to establish a
datum from which the vertical deviation of the pavement is measured. The pavement’s
vertical deviation is measured using a profile wheel, usually located at the midpoint of
the profilograph [50]. California and Reinhart profilographs are the most popular types
in North America [50]. The use of profilographs to collect data at the network level is not
popular [47,50].
Response-type equipment
Response-type equipment includes mechanical or electronic devices to evaluate pave-
ment roughness indirectly. These equipment pieces are calibrated by correlating the mea-
sured values with standardized roughness measures. They quantify the vehicle response
either mechanically (bump integrator) or using accelerometers [52]. Response-type equip-
ment was the primary technology for evaluating pavement roughness in the 1970s and
1980s [18,19,25].
Instead of measuring the actual road profile, response-type equipment measures the
vehicle response to the pavement roughness. Thus, this type of equipment is prone to
various sources of bias and uncertainties, including changes in vehicle dynamics and
equipment sensitivity. However, multiple benefits brought on by response-type equipment
made them the most popular roughness measurement equipment for several decades. The
advantages of the response-type equipment include low cost, ease of use, operation at high
speed, ability to collect large amounts of data, and the ease of correlating their measures
with PSI and standardized roughness values [19]. However, they are adversely criticized
because reproducibility is challenging to achieve, as the measurements are susceptible to
many factors, as mentioned before. Thus, frequent and rigorous calibration was required to
achieve any consistency in the collected data [19]. Conventional response-type equipment
such as Mays and Cox meters have been gradually phased out and replaced with more
precise and accurate measurement technologies.
Recently, this type of technology has been revived by utilizing smartphones, vehicles
with built-in sensors, and cheap accelerometers. Their application for measuring pavement
roughness has been the focus of enormous research efforts [54–61].
Inertial profilers
Inertial profilers, or simply profilers, are used to measure the actual pavement profile
and do not require a comparison with other equipment for calibrating the results. Instead,
it is necessary to calibrate the equipment’s sensors and associated electronics and ensure
the proper functioning of the computing hardware and software [50]. Inertial profilers are
essentially composed of accelerometers, height sensors, distance sensors, and computing
hardware and software [50]. Accelerometers are used to establish a reference plane from
which the vertical deviation of the vehicle is measured [50]. In the early generation of
pavement roughness measurement equipment with inertial referencing systems, the road
profile was defined using a wheel in contact with the pavement surface [62]. However, in
Remote Sens. 2023, 15, 2941 15 of 44

the latter generations, road profile has been established as the distance between the vehicle
and the pavement surface using non-contact height sensors, which can be ultrasonic,
infrared, or laser. The obtained road profile is adjusted by accounting for the vertical
movement of the vehicle captured using accelerometers. The true relative profile is then
used to calculate IRI by applying the quarter car simulation [10,63].
Inertial profilers with non-contact height sensors have become increasingly prevalent
since the 1990s [23,27]. Among other types, laser sensors have become the primary choice
in inertial profilers [23,26]. ASTM E950/E950M–09 [64] classified methods for measuring
pavement roughness using inertial profilers into four classes based on resolution and
precision. It is worth noting that the latest version of the ASTM E950/E950M–22 [65] did
not discuss such a classification.
However, laser-based inertial profilers have some limitations, including high initial
costs. Laser-based inertial profilers’ prices are much higher than response-type equipment
and manually operated profilers [52]. Moreover, they are typically inefficient at low
speeds [26,52]. Thus, highway agencies usually set a low-speed provision for operating
inertial profilers. For example, the Saskatchewan Ministry of Highways and Infrastructure
(MHI) requires a minimum collection speed of 25 km/h. Similarly, North Carolina and
Virginia DOT require data collection speeds greater than 24 km/h [26].
Three-dimensional imaging systems
Technology advancements in imaging systems have led to the development of auto-
mated tools capable of collecting different types of pavement condition data. The use of 3D
imaging techniques instead of 2D imaging constituted a colossal leap in the development
of automated systems. Three-dimensional imaging systems enable the efficient and highly
accurate collection of pavement condition data. Such technology can capture as-built
physical data in digital format to construct highly accurate 3D profiles of pavement [66]. Its
advantages include facilitating the evaluation of roughness at any track on the paved road.
Additionally, it facilitates surveying for various pavement condition data, including rough-
ness, rutting, cracking, and raveling [26]. In the early 2000s, several companies developed
3D imaging systems to evaluate pavement conditions, primarily roughness and rutting.
The developed systems utilized scanning lasers and reflectors to measure the reflection
times across the pavement surface to establish the 3D pavement profiles [23]. In the follow-
ing years, 3D imaging systems have been integrated into automated systems developed by
major vendors, including ARRB Systems (Rowville, Australia), Mandli Communications
(Fitchburg, WI, USA), and Pathway Services Inc. (Broken Arrow, OK, USA).
Different 3D imaging systems were explored for a variety of motivations. Several
studies [67–69] explored using low-cost depth cameras instead of the relatively expensive
laser profilers to evaluate pavement roughness. The use of 3D systems to construct the
road profile for evaluating road roughness is proposed to overcome the limitations of the
laser profiler at evaluating roughness at low and varying speeds [70]. Additionally, 3D data
can be used to assess other pavement condition aspects, including cracking, potholes, and
patching [68]. Additionally, such systems can be mounted on unmanned aerial vehicles
(UAVs) to reduce the time, costs, and efforts for roughness data collection [71].
Three-dimensional data can be obtained via different means, including stereography,
time-of-flight measurement, and triangulation. Stereography uses 2D pictures from a
moving camera or different cameras to construct 3D pavement profiles [72]. Stereography
is used in the Intel depth camera (D435i) [69]. Time-of-flight is used in Lidars as well
as Microsoft Kinect One depth camera [67]. Triangulation is widely used in commercial
profiling systems [9].

4.2.2. Roughness Measurement Technologies in Industry Practice

Period I: Until 1986
The state of the practice of pavement roughness measurement technologies is con-
cluded using two syntheses of Hicks and Mahoney [19] and Epps and Monismith [25].
 traveling over the pavement to give a subjective rating on a 0–10 scale [25].
Hicks and Mahoney [19] found that only the USA Army Construction Engineering
Research Laboratory (CERL) used a laser profiler to evaluate the pavement roughness [19].
However, the range and resolution of the early laser profilers were very limited. The laser
profiler used by CERL was operated at a slow speed of about 5 km/h [25]. Remarkably,
Remote Sens. 2023, 15, 2941 16 of 44
Epps and Monismith [25] reported that some agencies were already using ultrasonic pro-
filers (5), including Cox Ultrasonic and South Dakota Profiler (SDP). The study revealed
an increased interest in developing sensor-based non-contact profilers.
The two
The syntheses reported
study revealed that the use of
several different pavement
companies, roughness
including K. equipment,
J. Law Engineers includ-
(Ypsilanti,
ing various response-type equipment, profilographs, contact profilers, and
MI, USA), Cox and Sons (Colfax, CA, USA), PASCO Corporation (Tokyo, Japan), and Dy- non-contact
profilers.
natest The number
Consulting of pavement
(Ballerup, roughness
Denmark), measurement
have developed equipment
non-contact piecesThese
profilers. owned by
pro-
surveyed agencies, as reported in the 1986 synthesis [25], is shown in Figure 11. However,
filers utilize accelerometers and microprocessors, as well as ultrasonic or laser sensors.
response-type equipment was, by far, the most popular means for collecting roughness
One of the significant hurdles to the widespread use of profilometers was their high price
data in the mid-1980s. Equipment that performs direct profile measurement was less
compared to conventional response-type methods. The former was reported to cost about
widespread. However, non-contact profilers were receiving considerable attention by the
fifteen times more than the latter [23].
mid-1980s.

Figure 11. Numbers of roughness measurement equipment pieces owned by surveyed agencies as
reported
reported in
in the
the 1986
1986 synthesis [25].
synthesis [25].

Period II: 1987–1994

Hicks and Mahoney [19] reported that seven out of the nine surveyed agencies used
response-type
Figure 12 equipment.
displays theFive years of
number later, Epps and
pavement Monismith
roughness [25] reported
measurement that such
equipment
equipment was still the primary means of collecting pavement roughness data.
pieces owned by surveyed agencies, as reported in the 1994 synthesis [27]. The response- As shown
in Figure 11, the majority of the agencies (22) were using the Mays meter. Other
type equipment was still quite popular. A remarkable finding of the 1994 synthesis [27] response-
based
was theequipment, suchnon-contact
surge in using as BPR and profilers.
PCA-type meters, was used by at least nine agencies.
Figure 11 shows that using profilographs
As shown in Figure 12, Mays, PCA, and Cox and ride
contact profilers
meters werewasusedlimited for large-
by twenty-two
scale roughness data collection surveys [25]. The use of subjective
agencies. Additionally, ten agencies were using ARAN response-type equipment. Theevaluation of road
use
roughness was seldom. Only Ontario was reported to evaluate road roughness by raters
traveling over the pavement to give a subjective rating on a 0–10 scale [25].
Hicks and Mahoney [19] found that only the USA Army Construction Engineering
Research Laboratory (CERL) used a laser profiler to evaluate the pavement roughness [19].
However, the range and resolution of the early laser profilers were very limited. The laser
profiler used by CERL was operated at a slow speed of about 5 km/h [25]. Remarkably,
Epps and Monismith [25] reported that some agencies were already using ultrasonic
profilers (5), including Cox Ultrasonic and South Dakota Profiler (SDP). The study revealed
an increased interest in developing sensor-based non-contact profilers.
The study revealed that several companies, including K. J. Law Engineers (Ypsilanti,
MI, USA), Cox and Sons (Colfax, CA, USA), PASCO Corporation (Tokyo, Japan), and
Dynatest Consulting (Ballerup, Denmark), have developed non-contact profilers. These
profilers utilize accelerometers and microprocessors, as well as ultrasonic or laser sensors.
One of the significant hurdles to the widespread use of profilometers was their high price
compared to conventional response-type methods. The former was reported to cost about
fifteen times more than the latter [23].
Period II: 1987–1994
Figure 12 displays the number of pavement roughness measurement equipment pieces
owned by surveyed agencies, as reported in the 1994 synthesis [27]. The response-type
equipment was still quite popular. A remarkable finding of the 1994 synthesis [27] was the
surge in using non-contact profilers.
 eration equipment where optical sensors replaced ultrasonic sensors.
The era of non-contact profilers was evident, with many agencies transitioning from
response-type equipment to non-contact profilers. Up to ten agencies reported using more
than one type of equipment. For example, Pennsylvania, Oklahoma, and South Carolina
Remote Sens. 2023, 15, 2941 were using SDP in addition to the Mays ride meter. Similarly, West Virginia and Ohio
17 of 44
were using the K.J. Law non-contact profilers alongside the Mays profiler [27].

Figure 12.Numbers
Figure12. Numbers of of
roughness
roughness measurement equipment
measurement piecespieces
equipment ownedowned
by surveyed agencies agencies
by surveyed as as
reported
reportedininthe
the1994 synthesis
1994 [27].[27].
synthesis

As shown in Figure 12, Mays, PCA, and Cox ride meters were used by twenty-two
Period III: 1995–2009
agencies. Additionally, ten agencies were using ARAN response-type equipment. The use
The 2004
of contact synthesis
profilers indicated
was limited. Onlythat
threeall the surveyed
agencies were foundagencies
to usehad transitioned
the early contact to non
contactof
profiler profiling equipment.
GMR. Although The equipment
the synthesis utilizedthe
did not indicate oneuseofof
three types of height
profilographs for the sensors
large-scale collection
lasers, acoustics, orof roughness
infrared. Ofdata,
the 51it was noted that
agencies that many agencies
reported owned
the type of and used
technology used
them for calibration and project-level data collection [27,50].
in their profilers, the majority (86%) used laser-based profilers. Laser sensors enable data
As shown
collection in Figureof
at intervals 12,25twenty-five
mm or less, agencies used of
at speeds SDP. SDP was developed
approximately 96 km/h in the
[23]. Other
early eighties by the South Dakota Department of Transportation. It consists of three
sensors of acoustics and infrared were less popular, with only three agencies reporting
subsystems to collect vehicle displacement using accelerometers, vehicle height using
using acoustic sensors, and four reporting using infrared sensors. This shift in sensor tech
ultrasonic sensors, and vehicle distance [63]. Other non-contact profilers included K. J. Law
nology
8300 andisCox
significant comparedUltrasonic-based
CS 8000 Ultrasonic. to the previousprofilers
synthesis [27],
were theconducted
most commonteninyears
the earlier
where most non-contact profilers comprised acoustic sensors. Laser-based
eighties and early nineties. K. J. Law Model 8300 profiler was among the newer generation profilers con
tinued to gain
equipment wherepopularity, while
optical sensors the use
replaced of acoustics
ultrasonic sensors.and infrared sensors was entirely
marginalized
The era of by the 2009profilers
non-contact synthesis
was [37].
evident, with many agencies transitioning from
response-type equipment of
The development to 3D
non-contact
imagingprofilers.
systemsUp to ten agencies
capable reported
of measuring using more
pavement roughness
than one type of equipment. For example, Pennsylvania, Oklahoma,
was noted in the 2004 synthesis [23]. At least two companies (Phoenix Scientific and South CarolinaInc. (San
were using SDP in addition to the Mays ride meter. Similarly, West Virginia and Ohio were
Marcos, CA, USA) and GIE Technologies Inc. (New York, NY, USA)) have reported the
using the K.J. Law non-contact profilers alongside the Mays profiler [27].
development of laser-based 3D imaging systems for collecting pavement roughness and
Period
ruttingIII:
data1995–2009
[23]. The interest in 3D profiling systems experienced a continuous rise in the
following years.
The 2004 synthesis indicated that all the surveyed agencies had transitioned to non-
contact profiling equipment. The equipment utilized one of three types of height sensors:
lasers, acoustics, or infrared. Of the 51 agencies that reported the type of technology used
in their profilers, the majority (86%) used laser-based profilers. Laser sensors enable data
collection at intervals of 25 mm or less, at speeds of approximately 96 km/h [23]. Other
sensors of acoustics and infrared were less popular, with only three agencies reporting
using acoustic sensors, and four reporting using infrared sensors. This shift in sensor
technology is significant compared to the previous synthesis [27], conducted ten years
earlier, where most non-contact profilers comprised acoustic sensors. Laser-based profilers
continued to gain popularity, while the use of acoustics and infrared sensors was entirely
marginalized by the 2009 synthesis [37].
The development of 3D imaging systems capable of measuring pavement roughness
was noted in the 2004 synthesis [23]. At least two companies (Phoenix Scientific Inc. (San
Marcos, CA, USA) and GIE Technologies Inc. (New York, NY, USA)) have reported the
development of laser-based 3D imaging systems for collecting pavement roughness and
rutting data [23]. The interest in 3D profiling systems experienced a continuous rise in the
following years.
Remote Sens. 2023, 15, x FOR PEER REVIEW 19 of 45
Remote Sens. 2023, 15, 2941 18 of 44

Period IV: 2010–2022

Period TheIV:use
2010–2022
of laser-based systems to collect roughness data has been increasingly prev-
alentThein North
use ofAmerica
laser-basedandsystems
other parts of the
to collect world. Pierce
roughness andbeen
data has Weitzel [26] and preva-
increasingly Pierce
and in
lent Stolte
North [2]America
indicated andthe popularity
other of laser-based
parts of the world. Pierce systems for collecting
and Weitzel [26] and roughness
Pierce and
data. The
Stolte systems used
[2] indicated include point
the popularity laser profilers
of laser-based andfor
systems 3Dcollecting
profilingroughness
systems. The
data.latter
The
systems usedwere include
hailed for collecting
point high-resolution
laser profilers continuous
and 3D profiling profiles
systems. Thetolatter
construct the
systems
were hailed
pavement for collecting
profile. They canhigh-resolution
obtain 2D data continuous
by capturingprofiles to construct
the intensity the pavement
of reflected light and
profile.
3D dataThey can obtain
by capturing the2D data or
range bydepth
capturing
[26]. the
Theintensity of reflected
3D profiling systemslight
are and 3D data
increasingly
by
usedcapturing
to develop thefully
range or depth [26].
automated The 3D
pavement profilingsystems.
inspection systemsTheyare increasingly used to
facilitate surveying
develop fullypavement
for multiple automatedcondition
pavementdata,inspection
includingsystems. They facilitate
roughness, surveyingand
rutting, cracking, for multi-
fault-
ple pavement condition data, including roughness, rutting, cracking, and
ing [2,26]. However, Pierce and Weitzel [26] concluded that additional research efforts are faulting [2,26].
However,
necessary Pierce and Weitzel
to validate [26] concluded
the accuracy that additional
and repeatability of the 3Dresearch efforts
profiling are necessary
systems for meas-
to
uring road roughness. The synthesis marked the need for a standardized method forroad
validate the accuracy and repeatability of the 3D profiling systems for measuring cal-
roughness.
culating IRIThe using synthesis marked
3D profiling the need for a standardized method for calculating IRI
systems.
using 3D profiling systems.
4.2.3. Roughness Measurement Technologies in Journal Articles
4.2.3. Roughness Measurement Technologies in Journal Articles
Different technologies have been explored for pavement roughness measurement in
Different technologies have been explored for pavement roughness measurement in
the last few decades. Figure 13 shows the popularity of different roughness measurement
the last few decades. Figure 13 shows the popularity of different roughness measurement
technologies among the research community throughout the four studied periods. During
technologies among the research community throughout the four studied periods. During
the 1980s and the early 1990s, most of the research efforts focused on conventional re-
the 1980s and the early 1990s, most of the research efforts focused on conventional response-
sponse-type methods. During the following decades, more diverse technologies were in-
type methods. During the following decades, more diverse technologies were investigated.
vestigated. Research exploring response-type techniques was revived, but using new
Research exploring response-type techniques was revived, but using new sensing tools,
sensing tools, i.e., smartphones, vehicles’ built-in sensors, and cheap mounted sensors.
i.e., smartphones, vehicles’ built-in sensors, and cheap mounted sensors. Laser-based
Laser-based profiling equipment was continually explored, though not as a leading re-
profiling equipment was continually explored, though not as a leading research direction.
search direction. Three-dimensional imaging systems have been explored during the third
Three-dimensional imaging systems have been explored during the third period. However,
period. However, they have only become a trending research topic in recent years. Re-
they have only become a trending research topic in recent years. Recently, tire-mounted
cently, tire-mounted sensors as well as satellite and airborne synthetic aperture radar
sensors as well as satellite and airborne synthetic aperture radar (SAR) images have been
(SAR) imagesinvestigated.
increasingly have been increasingly investigated.

Figure 13.
Figure 13. Popularity
Popularity of
of major
major roughness
roughness data
data collection
collection technologies
technologies in
in the
the research
research community.
community.

Period I: Until 1986

A total
A total of
of fourteen
fourteen articles
articles published
published between
between 19711971 and
and 1986
1986 focused
focused onon studying,
studying,
developing, calibrating,
developing, calibrating,andandcomparing
comparingdifferent
different ride
ride quality
quality andand roughness
roughness measure-
measurement
ment methods were analyzed. Eleven of the fourteen retrieved articles
methods were analyzed. Eleven of the fourteen retrieved articles were published between were published
between
1983 and 1983
1986,and 1986, indicating
indicating a noticeablea noticeable
interest in interest
pavement in roughness
pavement roughness
measurement measure-
by the
ment by community
research the researchincommunity in the
the mid-1980s. mid-1980s.
Table 6 presentsTable
the6used
presents the used
roughness roughness
measurement
measurement
methods. methods.inAs
As explained explained
Table 6, mostin Table (10)
studies 6, most studies
focused (10) focused onequipment,
on response-based response-
based equipment,
particularly the Maysparticularly
meter. The thesame
Maystrend
meter. of The
usingsame trend of using
response-type response-type
equipment (Mays
equipment
meter and a(Mays meter and
fifth wheel bump a fifth wheel bump
integrator) integrator)
was also was
vivid in thealso vivid inSome
industry. the industry.
studies
Some studies experimented
experimented withprofilers
with non-contact non-contact profilers [73,74].
[73,74].
Remote Sens. 2023, 15, 2941 19 of 44

Table 6. Roughness measurement approaches used in articles published in study period I.

Roughness Measurement Approach Number of Studies *

Vehicle response 10
Panel rating 4
Contact profiler 2
Non-contact profiler 3
* Some studies made use of multiple approaches.

Period II: 1987–1994:

Ten research articles were retrieved in the period that extends between the two synthe-
ses of [25,27]. Eight of the ten articles were published between 1987 and 1989. This indicates
a relatively high interest in developing, calibrating, and correlating different equipment
and approaches for measuring pavement roughness in the 1980s compared to the early
1990s. Table 7 summarizes the used roughness measurement approach in articles published
in study period II.

Table 7. Roughness measurement approach used in articles published in study period II.

Roughness Measurement Approach Number of Studies *

Response-type equipment 6
Panel rating 3
Profilographs 1
Non-contact profilometer 3
* Some studies made use of multiple approaches.

Reflecting the industry’s state of the practice at the time, most articles (6) focused
on response-type equipment. About a third of the articles addressed the calibration of
response-type equipment. Two articles reviewed the non-contact profilers of SDP [63] and
PRORUT [75]. The correlation of profilographs [76] and Mays meter [77] measurements with
ride quality was also addressed. The use of non-contact profilers, including the K.J. Law
profiler, to calibrate the Mays meter was also investigated [78]. Only one study suggested a
new hardware development by developing a response-type equipment using a transducer
only to simplify the device [79]. Noticeably, Huft et al. [63] reported a study of testing and
improving the SDP in collaboration with the South Dakota Department of Transportation.
Period III: 1995–2009
Little research output (six articles) was retrieved on investigating, development, and
correlation of pavement roughness equipment between 1995 and 2009. Table 8 presents the
various roughness measurement approaches explored in the analyzed articles. Three articles
were published between 1995 and 2004. Lenngren [80] reported the successful experience
of The Royal Swedish Fortifications Administration in using a laser-based system called
Laser Road Surface Tester (LRST) to measure the roughness of airfield pavement. Another
study investigated a response-type system named DYNVIA [81]. Ahlin et al. [82] compared
the pavement roughness measured using a laser-based profiler, which measured based on
truck wheel vibration using accelerometers [82].
Table 8. Roughness measurement approach used in articles published in study period III.

Roughness Measurement Approach Number of Studies *

3D laser scanner 1
Laser profiler 4
Vehicle response 3
* Some studies made use of multiple approaches.
Remote Sens. 2023, 15, x FOR PEER REVIEW 21 of 45

Remote Sens. 2023, 15, 2941 20 of 44

or those available in vehicles and smartphones will significantly surge in the following
years.

TableBetween 2006measurement
8. Roughness and 2008, three articles
approach were
used also retrieved.
in articles publishedChang et period
in study al. [66]III.
conducted
their investigation using a 3D laser scanner for obtaining road profiles. The new approach
Roughness
was compared with anMeasurement Approach
inertial profiler (Multiple Laser Profiler), Number
as wellofasStudies
rod and* level.
Roughness measures3D laser scanner
obtained based on the 3D laser scanner and the multiple 1 laser profiler
were found to be highlyLaser profilerChen et al. [83] studied the harmonization
correlated. 4 of an inertial
profiler and a three-meter straight
Vehicle response edge. González et al. [39] attached an 3accelerometer to
*aSome
vehicle to evaluate
studies made usepavement
of multipleroughness.
approaches. The use of attached accelerometers or those
available in vehicles and smartphones will significantly surge in the following years.
Period
Period IV:
IV: 2010–2022
2010–2022
There has been a remarkable surge in research exploring alternative pavement rough-
There has been a remarkable surge in research exploring alternative pavement rough-
ness
ness measurementapproaches
measurement approachesininthe past
the few
past years.
few Of the
years. Of 130
the retrieved articles,
130 retrieved 100 were
articles, 100
published between 2010 and 2022. These articles mainly focused on
were published between 2010 and 2022. These articles mainly focused on developingdeveloping new ap-
proaches to increase
new approaches cost-effectiveness,
to increase meet themeet
cost-effectiveness, growing demanddemand
the growing for collecting rough-
for collecting
ness data from new industries, and leverage technological advancements and
roughness data from new industries, and leverage technological advancements and trends, trends, such
as crowdsourcing. Figure 14 shows the various approaches explored for
such as crowdsourcing. Figure 14 shows the various approaches explored for measuring measuring pave-
ment roughness.
pavement As depicted
roughness. in the
As depicted infigure, researchers
the figure, havehave
researchers explored a diverse
explored a diverserange of
range
technologies for this purpose. However, utilizing measured vehicle response
of technologies for this purpose. However, utilizing measured vehicle response through through
smartphones,
smartphones, vehicles’
vehicles’ built-in
built-in sensors,
sensors, andand mounted
mounted sensors
sensors has
has emerged
emerged as as the
the most
most
prominent
prominent trend.
trend. These
These techniques
techniques areare comparable
comparable to to the
the conventional
conventional World
World Bank
Bank Class
Class
III
III roughness
roughness measurement
measurement equipment
equipment [84].
[84].

Figure
Figure 14.
14. Roughness
Roughness measurement
measurement approach
approach used
used in
in articles
articles published
published in
in study
study period
period IV.
IV.

Acceleration-Based Response-type Approaches

Acceleration-Based
The
The central
central theme
theme of of research
research in
in this
this period
period is
is the
the use
use of accelerometers
accelerometers to to measure
measure
vehicle response
response for
forpavement
pavementroughness
roughness evaluation.
evaluation. TheThe acceleration
acceleration signals
signals are col-
are collected
lected using smartphones,
using smartphones, mounted mounted
sensors,sensors, and vehicles’
and vehicles’ built-inbuilt-in
sensorssensors or simulated
or simulated using
using
specialspecial software.
software. Most ofMost of theutilized
the studies studiesfour-wheeled
utilized four-wheeled vehiclessmartphone
vehicles to collect to collect
acceleration acceleration
smartphone data. The most commonly
data. The mostused vehiclesused
commonly are passenger
vehicles arevehicles (PV),vehicles
passenger such as
sedans,
(PV), compact
such cars,compact
as sedans, sport utility
cars,vehicles, and pick-up
sport utility vehicles,trucks [56,84–87].
and pick-up Some
trucks studies
[56,84–87].
considered
Some studiesbuses [88], whereas
considered otherwhereas
buses [88], studies other
used astudies
combination of PV and buses
used a combination [89]
of PV andas
well as
buses PVasand
[89] wellheavy-duty trucks (HDT)
as PV and heavy-duty [90,91].
trucks (HDT)Only a fewOnly
[90,91]. studies considered
a few using
studies consid-
two-wheeled
ered vehicles of vehicles
using two-wheeled cycles [41]
of and
cyclesmotorcycles [57].
[41] and motorcycles [57].
Smartphones have been increasingly explored as a cost-effective tool for evaluating
pavement roughness. The capabilities of smartphones and their widespread use make
them an attractive solution. Smartphones are equipped with a variety of sensors, including
Remote Sens. 2023, 15, 2941 21 of 44

GPS and accelerometers. The smartphones’ built-in sensors can be leveraged to collect
data required for pavement roughness evaluation. Additionally, smartphones support the
transference of the collected data readily through Wi-Fi or mobile data. Furthermore, the
use of smartphones for data collection is handy and does not require specialized personnel.
Smartphones measure roughness based on the vehicle response rather than the direct
pavement profile. Thus, multiple variables can affect the collected data, including the
smartphone type, vehicle model, driving velocity, mounting arrangement, and sampling
frequency. Table 9 lists the studies that explored the effect of the main factors affecting
pavement roughness evaluation using smartphones.
Table 9. Major studied attributes affecting pavement roughness evaluation using smartphones.

Attributes Articles
Smartphone type [56,85,90,92–95]
Vehicle type [85,89–91,94,96,97]
Mounting arrangement [55,89,90,94,95,97]
Speed [55,89,90,94,95,97–99]
Sampling rate [55,100,101]

Smartphones measure acceleration using the embedded micro-electromechanical

systems inertial motion unit (MEMS IMU). Thus, the measured acceleration is expected to
vary depending on the embedded MEMS IMU. Different smartphones have MEMS IMU
with distinct hardware, antenna size, and installation location. For example, variation in
the embedded accelerometers can affect the reliability of the data and bias the measured
roughness. Accelerometers with a resolution of 10 mm/s2 introduce double the inherent
noise compared to that having a resolution of 1 mm/s2 [56]. The effects of smartphone type
on pavement roughness measurement have been widely studied, with multiple studies
reporting a significant impact [56,85,90,93–95]. For example, Zhang et al. [92] reported a
relative error of up to 11%, which they considered tolerable.
Variations in the vehicles’ dynamic systems lead to differences in the filtering of
vibration caused by road roughness [95]. Most studies reported a significant impact
of vehicle model variation on the measured pavement roughness [85,89,90,94–96]. For
example, Hanson et al. [94] reported that the difference between the measured IRI and the
reference value varied from 5.4% to −11.9% when measured using a compact car (2001
Pontiac Sunfire) and a sport utility vehicle (2011 Nissan Rogue), respectively. However,
Botshekan et al. [97] indicated that well-trained machine learning models could be used
for different types of vehicles. Additionally, Jeong et al. [102] suggested that the use of
gyration (i.e., angular velocity) can minimize the effect of vehicle model variation.
Vehicle speed was reported to have a remarkable effect in the majority of the
studies [55,89,90,95,98,99]. It was remarked that higher speed results in an increased excita-
tion in the vehicle suspension. Thus greater amplitude and acceleration are observed [95].
For example, Bisconsini et al. [95] experimented with speeds of 40, 60, and 80 km/h as well
as a speed varying between 30 and 120 km/h. The study reported the lowest correlation
between the smartphone-based and reference roughness measures (R2 = 0.65) at varying
speeds. The highest correlation (R2 = 0.86) was found at a 60 km/h speed. However, it is
crucial to indicate that effect of the speed variation depends on several factors, including
the range of speed variation, the types of wavelengths present in the pavement profile, and
the vehicle’s dynamic system [94,95]. However, Hanson et al. [94] reported a limited impact
of speed change from 50 to 80 km/h in their study. Additionally, Botshekan et al. [97]
reported a lower impact of speed variation for larger-scale surveys.
The ideal scenario is to collect the chassis response directly. However, this is not
practical when using smartphones. Moreover, the mounting arrangement may differ sig-
nificantly under normal usage conditions when using smartphones as a crowdsourcing
tool. The mounting arrangement was found to have a considerable effect on the measured
Remote Sens. 2023, 15, 2941 22 of 44

roughness [55,89,90,94]. For example, Hanson et al. [94] found that the vent mounting
arrangement increased the relative error of the measured roughness to about 86% compared
to 1% for the best performing arrangement of windshield mounting. However, Biscon-
sini et al. [95] concluded a limited impact of the mounting arrangement. In general, a
rigid mounting arrangement is preferable to acquire the chassis vibration as accurately as
possible [95,97].
The sampling frequency is another important factor that can affect the evaluation
of pavement roughness. A higher sampling frequency is better for improving the signal
quality as well as the consistency and reliability of the results [56,93,101]. Some studies
used sampling rates as high as 400 Hz [56], whereas others explored sampling rates as
low as 16 Hz [57]. Multiple studies used a sampling frequency of 100 Hz [41,99,102].
Alatoom and Obaidat [55] investigated four sampling rates of 25, 50, 100, and 200 Hz.
The best performance was reported as the highest sampling rate of 200 Hz. Additionally,
Zeng et al. [101] indicated that fewer runs are required to achieve results at a tolerance
error rate of 10% at an acceptance confidence level of 95% at a higher sampling rate.
Bridgelall [100] reported a similar trend after comparing eight sampling rates.
Results inconsistency in using smartphones is a major limitation [92]. To improve
the accuracy and consistency of the results, Opara et al. [103] and Zhang [92] suggested
averaging the IRI values calculated using different smartphones. Moreover, it was fre-
quently reported that increasing the number of samples helps overcome this limitation.
Bisconsini et al. [95] illustrated that increasing the number of traversals from five to ten
reduced the coefficient of variation from 10–20% to 4–9%. Bridgelall [88] reported that
using 30 traversals from 18 buses reduced the margin of error to less than 6%. Similarly,
Ahmed et al. [93] suggested having 24 and 35 traversals for paved and unpaved roads,
respectively. Medina et al. [90] indicated the need for 50 to 60 traversals. Improving consis-
tency can be achieved by utilizing smartphones as a crowdsourcing tool. In fact, several
articles promoted the use of smartphones as crowdsourcing tools and suggested analyzing
a combination of data from different sources [57,90,91].
Two articles developed ML models to predict road roughness using signals obtained
using smartphones. Jeong et al. [102] developed a CNN model to calculate IRI using images
representing the raw time–history measurement of the vehicle response. Gamage [104]
developed an extreme gradient boosting to calculate IRI using features describing the
obtained signal: the mean, maximum, and standard deviation of acceleration signals, the
number of spikes in x, y, and z directions, as well as the average speed of the vehicle.
Mounted accelerometers that constitute a more similar approach to the early response-
type methods have been investigated in combination with different types of sensors. Many
studies have explored using accelerometers mounted to vehicles to measure pavement
roughness. Attached accelerometers can better acquire the vibrations of vehicles’ chassis
by reducing the variability caused due to smartphone mounting. However, compared to
smartphones, their use is more challenging due to the need for additional sensors such as
GPS and special arrangements to collect, store and transfer the data.
Nineteen articles published during the fourth study period were analyzed. Most
studies used field tests for data collection and validation, whereas only two used testbeds.
Some studies used simulated data alongside real field testing [105–107]. Additionally, most
of the studies explored the effect of vehicle speed, whereas only a few articles investigated
the impact of vehicle type, mounting arrangement, and sampling rate on the measured
roughness, as presented in Table 10.
Similar to the discussion made for smartphones, speed was reported to have a signifi-
cant impact on the measured pavement roughness [44,61,106–110]. It was noted that the
higher the speed, the greater the excitation of the vehicle’s suspension, resulting in signal
amplification [61]. For example, Chen and Xue [107] developed a road roughness level
identification model based on the BiGRU network. The model’s overall accuracy dropped
drastically from about 95% to less than 30% when tested on data collected at speeds of 20
and 80 km/h, respectively. However, Li et al. [111] utilized the inverse pseudo-excitation
Remote Sens. 2023, 15, 2941 23 of 44

method (IPEM) to calculate IRI by measuring vehicle suspension. Li et al. [111] reported
limited variation of measured roughness due to speed variation (40–120 km/h) using his
model. Moreover, some publications suggested corrections of the measured acceleration or
roughness considering the variation in speed [51,112].
Li et al. [111] reported a significant impact of the vehicle type on measured roughness.
Mounting arrangement and sampling rates were reported to impact the measured roughness
as well. Li et al. [108] reported a significant increase when lowering the sampling frequency.
The absolute error doubled when the sampling frequency was downgraded from 200 to
25 Hz. The use of sampling rates lower than 100 Hz was discouraged to avoid unreliable
results [108]. Chou et al. [61] found that placing the sensor on the dashboard gave better
results compared to the floor of the passenger seat and the floor of the central seat.
Table 10. Major studied attributes affecting pavement roughness evaluation using attached accelerometers.

Attributes Articles
Vehicle Type [111]
Mounting arrangement [61]
Speed [44,61,106,107,109–111,113]
Sampling rate [108]

Chen and Xue [107] developed an ML model using a bidirectional gated recurrent unit
(BiGRU) network. The model used the acceleration of the vehicle as well as the wheel to
classify the roughness level of the road. Liu et al. [105] developed an improved restricted
Boltzmann machine deep neural network algorithm based on the Adaboost Backward
Propagation algorithm. The input of the model was the vehicle acceleration and pitch
acceleration, whereas the output was the road roughness class.
Built-in sensors available in connected cars are one of the promising approaches to eval-
uating pavement roughness [114]. It offers an alternative to using specially equipped vehicles
and skilled technicians. Compared to smartphones, it minimizes the number of variables
affecting the collected signal, particularly eliminating the variability of smartphone type and
mounting arrangement. In addition, it allows for the utilization the known information of
the vehicle type. Six studies explored the possibility of using vehicles’ built-in sensors for
pavement roughness measurement [45,54,58,59,115,116]. Most studies used simulated data in
addition to the data collected from vehicles’ built-in sensors [45,58,59,115]. Mahlberg et al. [54]
explored the pertinence of using crowdsourced data from connected vehicles for pavement
roughness evaluation. Moreover, Liu et al. [45] demonstrated the applicability of using
crowdsourcing and semi-supervised learning to evaluate pavement roughness.
Nitsche et al. [115] developed three models using ANN, SVM, and random forests
to predict road roughness. Thirty-five features derived utilizing the output of the vehicle
response simulation were used to calculate the weighted longitudinal profile. The random
forests model was reported to demonstrate the best performance. Interestingly, Liu et al. [45]
developed a semi-supervised learning model to work with data coming from vehicles of
known types. The model utilizes a set of data obtained for roads of known roughness and
unknown vehicle type and a smaller set of data from known vehicles passing over roads of
known roughness. The model can work to predict both the vehicle parameters for vehicles
of unknown type and the roughness.
Simulated acceleration data representing vehicle vibration due to surface irregular-
ities that has been investigated to assess pavement roughness. Thirteen studies used
simulated acceleration data exclusively to develop and test road roughness evaluation.
Multiple studies used specialized simulation software, including ADAMS software (2) and
CarSim (1). They were used to simulate the interaction between pavement and vehicles.
Most studies (9) implemented their models using MATLAB software. Studies developed
simulation models of different degrees of freedom (DOF), including two (6), three, six, and
nine (1), four (3), and seven (3).
Remote Sens. 2023, 15, 2941 24 of 44

Seven articles used ML algorithms in their studies. Types, input, and output of the
used ML models are shown in Table 11. As presented in Table 11, different variations
of the neural network were used in the seven studies. Wang et al. [117] found that the
performance of ANN was comparable to the random forest and better than that of the
decision regression tree. Most of the articles developed supervised ML regression models.
Five articles developed ML models to predict quantitative values of pavement roughness.
One article developed an ML model to classify pavement roughness, considering five
roughness classes. Additionally, one study developed both classification and regression
ML models. Inputs of the different ML include features describing vehicle suspension,
including vertical acceleration. Wang et al. [117] demonstrated the importance of including
additional features, including vehicle type and speed, to boost the models’ performance.
Table 11. ML models developed using simulated data.

Study ML Input Output

ANN 76 features (acceleration signal) IRI
[114]
ANN 9 features (acceleration signal) Roughness class (1–5)
ADV (acceleration divided by velocity in a time
[118] ANN Roughness class (1–5)
series considering a time window of 7.5 s)
Acceleration of a vehicle, delayed version of the
[119] ANN Pavement roughness (mm)
vehicle acceleration and the road profiles
Grms (RMS body acceleration), speed, suspension
Decision regression tree,
[117] stiffness, suspension damping, tire pressure, sprung IRI
Random forest, ANN
mass (number of people)
general regression neural Vertical acceleration and the pitching angle
[120] Pavement roughness (mm)
network (GRNN) acceleration of the vehicle mass center

3D Imaging Systems
Multiple studies have used 3D imaging systems for pavement roughness measurement.
Different 3D imaging systems were explored considering one of the three main motives:
cost reduction, operation at varying speeds, and development of integrated systems to
collect various condition data in addition to roughness. Several studies [35,36,113] explored
using low-cost depth cameras instead of the relatively expensive laser profilers to evaluate
pavement roughness. The use of 3D systems to construct the road profile for evaluating road
roughness is proposed to overcome the limitations of laser profiler at evaluating roughness
at low and varying speeds [114]. Additionally, 3D data can be used to assess other pavement
condition aspects, including cracking, potholes, and patching [113]. Additionally, such
systems can be mounted on unmanned aerial vehicles (UAVs) to reduce the time, costs,
and efforts required for roughness data collection [115].
Depth cameras were used in three studies published between 2015 and 2022 for assess-
ing pavement roughness. Specifications of the used depth cameras are presented in Table 12.
As presented in Table 12, the Intel depth camera (D435i) utilizes stereoscopic technology
using imagers, whereas the Microsoft Kinect One utilizes time-of-flight technology using
an infrared sensor and a camera. Aleadelat et al. [69] utilized a low-cost Intel depth camera
(D435i (Intel RealSense, California, USA)) to evaluate pavement roughness. After scanning
road pavement, a depth stream was extracted. Subsequently, the depth stream was filtered
to generate the road profile to ultimately evaluate pavement roughness as IRI values. The
proposed approach successfully classified 76% of the tested road segments into the correct
road performance category as per the FHWA IRI threshold. Mahmoudzadeh et al. [68]
and Khalifeh et al. [67] used RGB-D sensors (Microsoft Kinect One (Microsoft Corporation,
Redmond, WA, USA)) to capture 3D images. The developed technology was reported
cost-effectively with good precision and accuracy. However, the vertical measurement reso-
lution of both camera cameras is larger than 0.01 mm [69,121]. Thus, both are comparable
to class 4 profilers, according to ASTM E1950-2018 [64,69].
Remote Sens. 2023, 15, 2941 25 of 44

Table 12. Specifications of the used depth cameras.

Number of Depth Vertical Resolution

Brand Infrared Manufacturer
Imagers Technology (Lower Limit)
Depth Camera D435i 2 No Stereoscopic 0.2 in Intel RealSense
Kinect One (v2) 1 Yes Time-of-Flight 0.04 in Microsoft

Laser-based 3D imaging systems were also used for pavement roughness measure-
ment. Tran and Taweep [122] and Kumar et al. [123] used mobile laser scanner (MLS)
systems to construct 3D point-cloud data along the road. The main components of the
GT–4 and XP-1 systems used by Tran and Taweep [122] and Kumar et al. [123], respectively,
are presented in Table 13. GT–4 MLS system was developed by Aero Asahi Corporation
(Tokyo, Japan), whereas the XP-1 MLS system was produced at the National University of
Ireland Maynooth (NUIM) [122,123]. Both systems include laser scanners, GPS units, and
cameras. The systems can be mounted on vehicles and support data collection at different
speeds. Tran and Taweep [122] reported collecting data at an average 50 km/h speed. The
developed approaches were reported to be able to evaluate pavement roughness with high
accuracy. It can also be used to calculate the roughness at any track on the road. However,
this approach is more data-intensive compared to conventional point laser profilers.
Table 13. Specifications of GT–4 and XP-1 mobile laser scanning system.

Article Brand 3D Laser Scanner GPS Camera

Laser scanners RIEGL
[122] GT–4 system GNSS/IMU unit 360-degree camera
VQ450 (2 sets)
[123] XP-1 system [124] Riegl VQ-250 IXEA LandINS GPS/INS unit 6 Progressive scan cameras

Additionally, stationary laser scanners were used to obtain 3D point cloud to evaluate
pavement roughness. Alhasan et al. [125] used a Trimble CX 3D STLS system to assess the
pavement roughness of two pavement sections of only 105 m. Although both studies illus-
trated the ability of stationary laser scanners to accurately evaluate pavement roughness,
their use is not practical for large-scale data collection programs.
Other studies used line laser sensors to develop 3D imaging of pavement surfaces
for roughness evaluation. Lee et al. [126] and Guo et al. [70] developed line laser-based
systems to obtain the 3D profile of pavement surface to measure road roughness. Both
systems include two laser scanners. Other sensors and tools are used, including cameras
and DMI units (for spatial referencing). The proposed systems were claimed to support
roughness measurement at low and varying speeds with high accuracy. Lee et al. [126]
reported that the system is reliable for unpaved roads.
Abohamer et al. [127] evaluated pavement roughness by employing ANN and multi-
nomial logistic (MNL) regression models and classified pavement sections into different
roughness categories based on 3D images. Additionally, a convolutional neural network
(CNN) was used to predict IRI. The study used 3D images obtained from Louisiana Dot
and Development (LaDOTD) pavement management system (PMS) inventory. Compared
with IRI values from the inertial profiler, the CNN model predicted IRI with a coefficient of
determination (R2 ) of 0.985 and an average error of 5.9%.
Stereography was used by multiple studies to generate 3D images of pavement
profiles for roughness evaluation using 2D cameras. Botha and Schalk [128] used two
low-cost cameras and an inexpensive inertial navigation system (SPAN-CPT) to construct
road profiles for roughness evaluation. Prosser-Contreras et al. [71] developed a system
that constitutes a camera and GPS unit mounted on a UAV to evaluate pavement roughness.
The method was found to be cost and time efficient. However, it suffers from multiple
limitations that necessitate further research. One limitation is the requirement to have the
Remote Sens. 2023, 15, 2941 26 of 44

road be traffic-free. Another limitation is the vulnerability of the data quality to the weather
and lighting quality.
Synthetic Aperture Radar
The use of SAR technology for evaluating pavement roughness has been the focus
of multiple studies [24,129–135]. SAR data from X-band [132–135], L-band [24,131], and
C-band [130] was investigated for pavement roughness estimation. The data were collected
using various satellites such as Sentinel-1 [24,130], Sentinel-2 [132], Cosmo-SkyMed [129],
and Advanced Land Observing Satellite (ALOS) [131]. Moreover, different ML algorithms
were utilized to evaluate pavement roughness. Fiorentini [24] explored various ML models
to evaluate roughness and concluded the superiority of SVM. Bashar and Torres-Machi [130]
applied deep learning to estimate the IRI. The developed model predicted IRI values with
a mean absolute error of about 0.23 m/km (14.6 in/mile). Karimzadeh and Matsuoka [134]
used the SAR backscattering values to assess IRI. The study reported a loose correlation
with the reference values obtained using BumpRecorder software (BumpRecorder Co.,
Ltd., Tokyo, Japan) and a smartphone. However, multiple studies developed models to
classify the roads based on their roughness. Karimzadeh and Matsuoka [132] suggested a
qualitative assessment instead of quantifying pavement roughness due to insufficient image
resolution. Karimzadeh and Matsuoka [132], Suanpaga et al. [131], and Meyer et al. [129]
classified roads into two classes; roads with good and bad roughness.
Satellite data offer wide-scale coverage, which can minimize the time and cost required
for pavement roughness evaluation. However, the applicability of using satellite data for
pavement roughness evaluation is still debatable at the current stage. Thus, more research is
required before assuming the robustness of this technology with the low-resolution publicly
available satellite images. Alternatively, Babu et al. [133] and Rischioni et al. [135] used
high-resolution airborne polarimetric SAR images to evaluate pavement roughness. The
two studies utilized DLR’s airborne radar sensor F-SAR to acquire fully polarimetric X-band
datasets. The estimated roughness was highly correlated with reference values obtained
using a handheld laser scanner. The proposed method sounds more robust compared to
satellite SAR-based approaches.
Tire Pressure Sensors
Few studies explored attaching sensors to vehicle tires to measure pavement roughness.
(Zeng et al. [136] estimated the IRI using a tire pressure sensor. The study employed a
combination of ideal gas low and elastic contact models to derive a nonlinear relationship
between tire pressure and contact force. The results indicate reliable IRI values, especially at
low vehicle speeds, can be obtained. Zhao et al. [137] and Zhao and Wang [138] developed
a new method based on dynamic pressure sensors and axle accelerometers to evaluate
pavement roughness. The measured IRI values were comparable with those obtained using
a laser profiler. Differently, Yang et al. [139] employed machine learning algorithms to
classify road roughness levels based on data collected using an intelligent tire system with
a piezoelectric cable. The layout of the piezoelectric cable was optimized using a finite
element model.
Other Technologies
Other technologies have been explored for measuring pavement roughness, such as
mechanical profilers [140] and directional microphones (acoustic-based method). Becker
and Els [140] developed a mechanical profiler called Can-Can Profilometer that was sug-
gested to be useful for obtaining 3D data for research purposes. Additionally, Zhao et al. [141]
developed an acoustic-based method that utilizes five directional microphones mounted
on a vehicle to evaluate road roughness. The study developed an algorithm to assess
pavement roughness based on the principle that higher sound pressure is associated with
higher pavement roughness. However, the robustness of the model is questionable as the
used algorithm was built using roughness data less than 2 m/km.
Remote Sens. 2023, 15, 2941 27 of 44

In-pavement strain sensors (optical fiber sensors) have been explored on a limited
scope through laboratory and limited field experiments for pavement monitoring [142],
including roughness evaluation [42,143]. While the technology has shown considerable
potential, its practicality remains questionable due to the need for optical fiber infras-
tructures. The implementation of such infrastructures can be costly, and it may not be
practical for many pavement monitoring applications, especially in remote or hard-to-
reach areas. Nonetheless, the technology remains a promising avenue for future research
and development, and ongoing efforts are being made to improve its practicality and
cost-effectiveness.
Infrared sensors and ultrasonic sensors are cost-effective methods that hold potential
for small agencies with limited resources. These sensors have shown promise in various
pavement monitoring applications, including roughness evaluation, and can offer an
affordable alternative to more expensive measurement technologies. While they may not
provide the same level of accuracy as higher-end sensors, they can still offer valuable
insights into pavement condition and can be a useful tool for smaller agencies looking to
prioritize maintenance and rehabilitation efforts.

5. Current Trends, Research Gaps, and Future Directions

5.1. Industry Practices
The techniques used in practice for pavement roughness measurement have been
evolving over the years. The most apparent trend is the increasing popularity of laser-based
technologies. Laser-based inertial profilers have gained significant momentum in the last
three decades. Moreover, 3D imaging systems are increasingly used for pavement condition
surveys to collect different types of data, including roughness.
Thirty-four pieces of equipment from over twenty vendors across various countries
were reviewed to identify the state of the practice for pavement roughness measurement.
Details of the reviewed pieces of equipment are presented in Table 14. As revealed in
Table 14, half of the reviewed equipment (17) are profilers of one or more point laser sensors.
Two-point laser sensors constitute the minimum requirement to measure roughness in
both vehicle wheel tracks. However, a larger number of point laser sensors are commonly
used to acquire additional data, including rutting and geometric measurements. Line laser
scanning sensors are used in three pieces of equipment. Moreover, two vendors offer to
provide line laser sensors in place of the point laser sensors. Such equipment can provide
additional condition data, including rutting and geometrics.
The use of cost-effective technologies continues to secure a niche market, especially for
low-volume and urban roads. Six pieces of equipment used accelerometers to evaluate the
pavement roughness. Four of them are walking profilometers where the equipment is rolled
along at walking speed. The other two can be mounted on vehicles to collect roughness
data more efficiently, although with lower accuracy. Additionally, the use of smartphones
for roughness measurement is facilitated by the commercial of BumpRecorder [144].
The use of laser triangulation technology to develop 3D imaging is used by the
Canadian company of Pavemetrics to develop the Laser Crack Measurement System
(LCMS® -2). LCMS-2 can be used to virtually collect all pavement surface condition data.
The use of 3D imaging has been gaining significant momentum in the last few years [2].
LCMS-2 is integrated into multiple commercial systems, including Hawkeye 2000 [145],
Mandli collection system [146], IrisGO Pave [147], and IrisPRO Pave [148]. Other 3D
laser imaging systems include PaveVision3D Ultra [149] and ZOYON-RTM [150]. Three-
dimensional imaging systems are expected to achieve multiple benefits, including accuracy
enhancement. Additionally, such systems facilitate the collection of different condition
data, including roughness, rutting, cracking, faulting, spalling, and potholes [2].
Remote Sens. 2023, 15, 2941 28 of 44

Table 14. Details of the commercially available systems for measuring pavement roughness.

Equipment Vendor Country 2 Type Technology Reference

BumpRecorder BumpRecorder Co., Ltd. Japan III Smartphone [144]
Roadmaster Road Roughness
AL-Engineering Oy Finland III Accelerometer [151]
Meter
Roughometer 4 ARRB Group Australia III Accelerometer [152]
Walking Profiler G3 ARRB Group Australia I Accelerometer [153]
Zhuozhou Tianpeng Instrument
DAPPES-3 (Walking profiler) China I Accelerometer [154]
Manufacturing Co., Ltd.
SurPRO International Cybernetics USA I Accelerometer [155]
CS8800 Walking Profiler Surface Systems & Instruments USA I Accelerometer [156]
Road Surface Profiler MK III Dynatest Denmark I Point laser (up to 21) [157]
Road Surface Profiler IV Dynatest Denmark I Point laser (1–2) [158]
Nanjing T-Bota Scietech
Road Surface Profiler China I Point laser (1–17) [159]
Instruments & Equipment Co., Ltd.
PaveProf V2.0 PaveTesting Ltd. UK I Point laser (up to 40) [160]
Digital Laser Profiler ARRB Group Australia I Point laser (3) [161]
LaserProf Greenwood Engineering Denmark I Point laser (2) [162]
Greenwood Profiler Greenwood Engineering Denmark I Point laser (48) [163]
IrisGO Portable Profiler International Cybernetics USA I Point laser (1–3) [164]
IrisPRO Inertial Profiler International Cybernetics USA I Point laser (1–5) [165]
ARAN inertial profiler Fugro Roadware Canada I Point laser (2) [166]
CS9550 Portable Survey Profiler Surface Systems & Instruments USA II 1 Point laser (2–6) [167]
CS8600 Ultra-Light Profiler Surface Systems & Instruments USA I Point laser (1) [168]
ROMDAS Laser Profilometer Data Collection Ltd. New Zealand I Point laser (1–2) [169]
APTech’s EDGE Applied Pavement Technology Inc. USA I Point laser (2) [170]
RoadSTAR Austrian Institute of Technology Austria I Point laser (21) [171]
8300 High-Speed Inertial Road
Ames Engineering USA I Line/Point laser (1–2) [172]
Profiler
6300 Lightweight Profiler Ames Engineering USA I Line/Point laser (1–2) [173]
Pathway inertial profiler Pathway services USA I Line laser [174]
PPS-Plus Fraunhofer IPM Germany I Line laser [175]
Gocator 2342 LMI Technologies Canada I Line laser [176]
3D laser imaging
LCMS® -2 Pavemetrics Canada I [177]
system
3D laser imaging
Hawkeye 2000 ARRB Group Australia I [145]
system [LCMS® -2]
3D laser imaging
Mandli collection system Mandli Communications USA [146]
system [LCMS® -2]
3D laser imaging
IrisGO Pave International Cybernetics USA I [147]
system [LCMS® -2]
3D laser imaging
IrisPRO Pave International Cybernetics USA I [148]
system [LCMS® -2]
3D laser imaging
PaveVision3D Ultra WayLink Systems Co. USA I [149]
system
Wuhan Optics Valley Zoyon Science 3D laser imaging
ZOYON-RTM China I [150]
and Technology Co., Ltd. system
1 Vertical resolution < 0.01 in; 2 Headquarter location.

Figure 15 illustrates the anticipated drivers and future trends in the industry practice
of pavement roughness measurement. As depicted in Figure 15, laser profilers and 3D
Remote Sens. 2023, 15, 2941 29 of 44

laser imaging systems are expected to maintain their prevalence, especially for high-volume
roads, due to their unmatched efficiency, accuracy, and precision. Additionally, they are
highly automated techniques. In particular, 3D laser imaging systems provide additional
advantages, including the ability to survey multiple pavement condition data at varying
speeds. The commercialization of smartphones and vehicles’ built-in sensors for roughness
measurement is also projected to increase. This trend is driven by their cost-effectiveness
Remote Sens. 2023, 15, x FOR PEER REVIEW and
31 of 45
applicability in various industries, as shown in Figure 15. Highway agencies are expected to
increasingly adopt these techniques, particularly for urban and low-volume roads.

Figure 15. Anticipated

Figure15. Anticipated future
futuretrends
trendsin
inindustry
industrypractice
practicefor
forpavement
pavementroughness
roughnessmeasurement.
measurement.

It is worth
5.2. Research noting that the analysis of journal article sponsorships indicates a lack of
Community
industry support for the academic research community, with only one article sponsored by
The developed approaches for measuring pavement roughness differ in their work
a commercial company. This highlights a low level of coordination between industry and
principle, costs, accuracy,
academia, which can result practicality
in a waste offor large-scale
resources. Givendatathecollection,
increasingand soundness
interest as
in rough-
crowdsourcing tools, as presented in Table 15. Most of the investigated
ness measurement research, the industry can leverage exhausted efforts by collaborating technologies are
low-cost
with and applicable
the research for data
community. Onecollection at the network
way to achieve level, establishing
this is through as presentedpartnerships
in Table 15.
Most investigated approaches utilize sensors mounted or available
between industry and academia to work together on research projects. Such partnerships in standard pas-
senger
can cars,joint
involve where data resource
funding, can be and
collected at traffic
expertise sharing,speed. Remarkably,ofsmartphones,
and co-authoring publications.
mounted accelerometers, and vehicles’ built-in sensors are cost-effective
Furthermore, government funding agencies can provide incentives for industry-academia techniques for
roughness evaluation.
collaborations by offeringHowever,
funding their indirect approach
opportunities for jointtoprojects,
evaluating roughnesspartner-
encouraging and the
high variability that can be introduced by multiple
ships, and prioritizing proposals that involve collaborations. factors, including the type of vehicle,
may lead the accuracy to be questionable. For example, Wang et al. [117] indicated that
5.2. Research Community
the accuracy of a smartphone-based approach could be comparable to a World Bank Class
3 device if the mounting
The developed is rigidfor
approaches and the speedpavement
measuring is considered. Without
roughness consideration
differ for
in their work
the vehicle type and speed, the accuracy was reported to be similar
principle, costs, accuracy, practicality for large-scale data collection, and soundness asto World Bank Class
4 approach (visual
crowdsourcing inspection).
tools, as presentedAs an
in alternative
Table 15. Mostto calibration, smartphones
of the investigated and vehicles’
technologies are
built-in sensors are the obvious options for crowdsourcing.
low-cost and applicable for data collection at the network level, as presented in Table 15.
Some of the technologies can be considered premature for widespread deployment,
such as piezoelectric cable on a tire and tire pressure sensors. They hold potential for ap-
plications in highly autonomous vehicles. As technology continues to evolve, these emerg-
ing technologies could become increasingly relevant in the near future.
Laser profilers still generally offer a reasonable option for roughness measurement.
However, their limitation in urban roads and considerable cost urge researchers to con-
sider alternatives. Three-dimensional imaging systems hold more significant potential as
a multipurpose approach for evaluating a variety of pavement defects, including rough-
Remote Sens. 2023, 15, 2941 30 of 44

Table 15. Comparison of the developed approaches for measuring pavement roughness.

Technology Type Cost Accuracy Network Level Crowdsourcing

Smartphones Indirect L L Yes Yes
Attached accelerometers Indirect L L Yes No
Vehicles’ built-in sensors Indirect L L Yes Yes
Vehicle telematics Indirect L L Yes Yes
Piezoelectric cable on a tire Indirect L L Yes No
Tire pressure sensors Indirect L L Yes No
3D imaging systems (laser) Direct H H Yes No
3D imaging systems (low-cost depth camera) Direct L L-M Yes No
3D imaging systems (stereography) Direct L-M L-H Yes No
Ultrasonic sensors (low-cost) Direct L L Yes No
Infrared sensors (low-cost) Direct L L Yes No
In pavement sensors Indirect - M No No
Microphones Indirect M L Yes No
SAR (satellite and airborne) Indirect L L Yes No
Laser-based equipment Direct H H Yes No
Can-Can (mechanical) Direct M H No No
L: low; M: medium; H: high.

Most investigated approaches utilize sensors mounted or available in standard passen-

ger cars, where data can be collected at traffic speed. Remarkably, smartphones, mounted
accelerometers, and vehicles’ built-in sensors are cost-effective techniques for roughness
evaluation. However, their indirect approach to evaluating roughness and the high vari-
ability that can be introduced by multiple factors, including the type of vehicle, may lead
the accuracy to be questionable. For example, Wang et al. [117] indicated that the accuracy
of a smartphone-based approach could be comparable to a World Bank Class 3 device if the
mounting is rigid and the speed is considered. Without consideration for the vehicle type
and speed, the accuracy was reported to be similar to World Bank Class 4 approach (visual
inspection). As an alternative to calibration, smartphones and vehicles’ built-in sensors are
the obvious options for crowdsourcing.
Some of the technologies can be considered premature for widespread deployment,
such as piezoelectric cable on a tire and tire pressure sensors. They hold potential for
applications in highly autonomous vehicles. As technology continues to evolve, these
emerging technologies could become increasingly relevant in the near future.
Laser profilers still generally offer a reasonable option for roughness measurement.
However, their limitation in urban roads and considerable cost urge researchers to consider
alternatives. Three-dimensional imaging systems hold more significant potential as a multi-
purpose approach for evaluating a variety of pavement defects, including roughness. Their
versatility in terms of technology type, accuracy, and cost offers a range of options for dif-
ferent applications. For less important roads or smaller agencies with limited resources, the
use of low-cost depth cameras or stereography with low-cost cameras may be appropriate
and offer a viable option. Alternatively, 3D laser imaging systems provide the highest level
of accuracy and efficiency and can be used in a range of different circumstances. However,
their significant cost remains a hurdle for agencies with limited resources.
Remote sensing technologies, such as airborne and satellite synthetic aperture radar
(SAR) images, offer significant potential for various pavement inspection applications,
including roughness evaluation. While satellite imagery is generally more cost-effective,
airborne SAR images can offer higher resolution and greater accuracy. However, challenges
Remote Sens. 2023, 15, 2941 31 of 44

such as the presence of vehicles on the road, tree coverage, and processing difficulties,
including segmentation challenges, can limit the effectiveness of this approach in spe-
cific scenarios, particularly on urban roads [134,135]. Additionally, it was reported by
Fiorentini et al. [178] that roughness resulting from endogenous factors, such as traffic flow
and pavement materials, may not be detectable using SAR technology. As a result, the use
of remote sensing technologies for pavement inspection may need to be carefully evaluated
on a case-by-case basis.
Figure 16 shows the current research gaps and the anticipated future research themes
regarding the measurement of pavement roughness. Precision issues and calibration diffi-
culties associated with using smartphones and connected vehicles are expected to intensify
the use of crowdsourcing and machine learning. The growing demand for roughness mea-
surement from new industries will boost the exploration of using smartphones and vehicles’
built-in sensors. Developing efficient and cost-effective approaches is also expected to
continue gaining attention from researchers. Three-dimensional laser imaging systems are
expected to draw more attention as a technology capable of acquiring different pavement
condition data. The problem of lack of standardization in testing new methods in academia
Remote Sens. 2023, 15, x FOR PEER REVIEW 33 of 45
requires further focus. The next discussion lays out the anticipated future research directions
and their associations with the identified research gaps and different technologies.

Figure 16.
Figure 16. Anticipated
Anticipatedfuture trends
future among
trends the research
among community
the research regarding
community pavement
regarding rough-
pavement
ness measurement.
roughness measurement.

Crowdsourcing
5.2.1. Crowdsourcing
use of
The use of acceleration
acceleration data,
data, particularly
particularly from
from smartphones
smartphones and and vehicles’
vehicles’ built-in
built-in
sensors, for
sensors, forroughness
roughnesspavement
pavementinspection
inspectionisisperformed
performed following
following either
either thethe calibration
calibration or
or crowdsourcing
crowdsourcing schemes.
schemes. TheThe
firstfirst scheme
scheme entails
entails calibrating
calibrating the the measures
measures withwith rough-
roughness
values collected
ness values by aby
collected World Bank
a World BankClass 1 profiler.
Class 1 profiler.However,
However,asasdiscussed
discussedbefore,
before, this
this
approach has
approach has a major limitation regarding the limited consistency of the acceleration data.
Additionally,
Additionally, itit requires
requires data
data collection
collection byby driving
driving the
the complete
complete network
network using
using the
the same
same
arrangement usedused forfor calibration,
calibration,including
includingvehicle
vehicletype
typeandand speed.
speed. TheThe other
other scheme
scheme of-
fers a more practical alternative by involving road users to cheaply collect large volumes
of data. Crowdsourcing in this regard is facilitated by the public for monitoring pavement
conditions. Larger datasets were reported to marginalize the effect of the individual fac-
tors on the measured roughness, including vehicle type, smartphone model, mounting
arrangement, and driving speed [91,104]. Roughness measured via crowdsourcing can
Remote Sens. 2023, 15, 2941 32 of 44

offers a more practical alternative by involving road users to cheaply collect large volumes
of data. Crowdsourcing in this regard is facilitated by the public for monitoring pavement
conditions. Larger datasets were reported to marginalize the effect of the individual
factors on the measured roughness, including vehicle type, smartphone model, mounting
arrangement, and driving speed [91,104]. Roughness measured via crowdsourcing can
help users choose the best route considering the ride quality and users’ expenses.
Table 16 presents details of some studies that used acceleration data collected using smart-
phones and connected vehicles as a crowdsourcing tool to evaluate road roughness. Table 16
shows that most studies used smartphones as a crowdsourcing tool [57,85,90,97,99,104]. This
is predictable as road users widely use smartphones, and customized software can be easily
installed to collect and analyze the data. Among the developed software are iRoads X [104]
and RoadSR [41]. However, two studies explored using connected vehicles as crowdsourcing
tools to evaluate roughness data [45,54]. Additionally, the use of crowdsourcing is usually
coupled with ML [45,97,104]. The majority of the studies presented colored maps to indicate
the roughness level of road sections [57]. As presented in Table 16, the scale of some studies
was large enough to include data collected for hundreds of kilometers collected using tens of
cars and smartphones. SmartRoadSense database covers about 24,000 km of roads collected
using 432 different vehicles and 147 different types of smartphones.
Table 16. Crowdsourcing approaches for measuring pavement roughness.

Article Used Technology Smartphone Types Vehicle Types Database Size (km) Application
[99] Smartphone 147 432 24,000 SmartRoadSense
[104] smartphone 2 3 56 iRoads X
[41] Smartphone 1 1 (bicycle) 13.456 RoadSR
[57] Smartphone 5 10 100 Unnamed
[85] Smartphone 2 3 0.35 Server-based system
[90] Smartphone 33 45 3.2 NA
[97] Smartphone 2 15 47 NA
[54] Built-in sensors - NA 32 NA
[45] Built-in sensors - 1 200 NA

Crowdsourcing is a promising alternative that is cheap and easy to deploy for pave-
ment roughness measurement. A growing interest in this research theme is expected to
persist and foster in the coming years. Moreover, anomaly detection techniques [179] and
blockchain technology [180–182] can be explored to enhance the reliability of the collected
data and address privacy issues.

5.2.2. Low-Cost Solutions

Many researchers have devoted their work to developing low-cost solutions for evalu-
ating pavement roughness [51,67,69,94,109,183,184]. Table 17 presents some of the low-cost
approaches investigated by multiple studies. As presented in Table 17, Aleadelat et al. [69] and
Khalifeh et al. [67] used depth cameras as cheap as USD 200 to evaluate pavement roughness.
Arbabpour Bidgoli et al. [109] developed an attached accelerometer-based system with an
estimated cost of USD 50. A USD 47 system consisting of an infrared-based sensor and an
accelerometer was developed by Kheirati and Golroo [184]. Some studies explored develop-
ing systems as cheap as USD 20 (Behera et al. [183]) and USD 29 (Mirtabar et al. [51]) using
attached accelerometers and ultrasonic sensors, respectively. However, the majority of these
approaches require rigorous calibration that necessitates Word Bank Class I profilers.
Remote Sens. 2023, 15, 2941 33 of 44

Table 17. Low-cost approaches for measuring pavement roughness.

Article Technology Cost (USD)

[67] Depth camera 200
[69] Depth camera 200
[86] Smartphone 100
[109] Attached Sensors 50
[184] Infrared sensors 47
[51] Attached Sensors 29
[183] Ultrasonic sensors 20

The use of drones and satellites can provide another opportunity to reduce the cost
of pavement inspection [71,130]. Alternatively, multiple studies explored using available
devices to calculate pavement roughness. Many studies used smartphones that are avail-
able to road users [56,99,104]. Other studies used vehicle telematic [185] and vehicles’
built-in sensors [54,116]. As an alternative to calibration, several studies explored using
smartphones and vehicles’ built-in sensors as crowdsourcing tools [85,104].
In fact, the measurement of road roughness does not conceptually present an extremely
challenging problem. On the other hand, measuring equipment costs are significant [79].
In a study published in 2007, Word Bank Class I laser-based profilers were reported to
cost between USD 25,000 and USD 50,000 [52]. The authors asked several vendors to
provide quotations to evaluate the current prices of laser profiler systems. The prices of
such equipment are not lower nowadays based on four quotes collected by the authors
from multiple companies in Asia, Australia, and North America. Small municipalities and
highway agencies with limited resources may be unable to afford advanced laser profilers,
especially in low-income countries. Thus, this research theme is expected to continue in the
coming years.

5.2.3. Machine Learning

The use of machine learning for roughness measurement has experienced a remark-
able increase in the last few years. Table 18 presents the used machine learning and the
associated data collection technology. As shown in Table 18, many ML models were used to
evaluate roughness using acceleration signals obtained using smartphones [102,104], vehi-
cles’ built-in sensors [45,115], mounted sensors [105,107], and simulated data [114,117–120].
Additionally, Abohamer et al. [127] developed different ML models to evaluate pavement
roughness using 3D data.
Table 18. ML models used for pavement roughness evaluation.

Article ML Technology
[102] CNN
Vehicle response (smartphones)
[104] Extreme Gradient Boosting
[115] ANN, SVM, and Random Forests
Vehicle response (vehicles’ built-in sensors)
[45] Self-supervised Learning
[105] Deep Neural Network
Vehicle response (mounted sensors)
[107] BiGRU
[114,118,119] ANN
[117] Decision Regression Tree, Random Forest, ANN Vehicle response (simulated data)
[120] General Regression Neural Network (GRNN)
[127] ANN, MNL, CNN 3D imaging systems
[24,130] SVM, Deep Learning SAR images
Remote Sens. 2023, 15, 2941 34 of 44

As presented in Table 18, ANN and deep learning and frequently used to evaluate
roughness data. The use of models other than supervised learning may provide a tool to
advance the use of crowdsourcing tools and overcome challenges such as incomplete data.
The use of ML is expected to keep increasing in the coming years.
Machine learning momentum is expected to be fostered by the increasing use of crowd-
sourcing as an alternative to calibration for vehicle response methods. In this regard,
machine learning can facilitate the development of cheap and more robust roughness evalu-
ation methods. Additionally, the use of machine learning models coupled with computer
vision algorithms is expected to rise with the growing interest in 3D laser imaging systems.

5.2.4. Growing Demand for Roughness Data Measurement by Different Industries

Information about pavement roughness is no longer the sole interest of highway
agencies. Instead, road users are becoming increasingly interested in understanding the
road’s condition. As discussed before, roughness is considered a major indicator of riding
quality and significantly impacts users’ comfort and expenses. Recently, an increasing
number of studies have been exploring tools to help users collect, visualize, and share
information on road roughness. With the availability of cheap sensors, smartphones, and
connected cars, many studies have started exploring crowdsourcing to derive and provide
road users with the required knowledge about road roughness [57,85,104,186]. Many of
these studies used maps to help users perceive road quality. Crowdsourcing tools can
help users share the measured road roughness and their experienced ride quality and fuel
consumption at different speeds. These pieces of information can help users compare their
route choices, adjust their driving speed, and better plan their trips [112].
Another roughness-measurement-demanding industry is the automotive industry.
Roughness has long been a matter of interest for the automotive industry due to its effects
on vehicle suspension systems. However, two developments in the automotive industry
have fueled the interest in road roughness measurement. The first significant development
is the evolution of vehicle suspension systems from passive to semi-active and active
control. This change makes the knowledge of road roughness of great importance to
improve ride quality. The other development is the pursuit of increasing the level of driving
automation to develop autonomous vehicles. Thus, enhancing vehicles’ ability to evaluate
road profiles has recently been a focus of a tremendous amount of research [45,54,187–189].
Equipping vehicles with the ability to evaluate road roughness is essential to facilitate
effective road-preview-based suspension control systems in future vehicles. In such systems,
the dynamic characteristics of the vehicle suspension system are adjusted to improve the
vehicle chassis performance and ride quality in real-time by utilizing roughness information
of the road ahead [190,191]. Thus, the automotive industry aims to equip vehicles with
pavement roughness technologies, which is essential to facilitate effective road-preview-
based suspension control systems in future vehicles.
Therefore, cars are expected to collect and share condition data using built-in sensors
and software. Several researchers have explored evaluating pavement roughness using
vehicles’ built-in sensors [45,54,58,59,115,116]. This approach is relatively more reliable
than smartphones due to eliminating the impact of some variables, namely the smartphone
type and mounting arrangement. Additionally, data from similar vehicles can be used
collectively to evaluate pavement roughness. Moreover, connected cars can be used as a
crowdsourcing tool to collect large amounts of data cheaply and easily [45,54].

5.2.5. Integrated and Comprehensive Condition Assessment Tools

The use of laser-based 3D imaging systems has been gaining increasing popularity in
recent years due to multiple factors, including high accuracy and precision, independence
from natural light, and applicability to urban roads. Additionally, it facilitates surveying for
various pavement condition data, including roughness, rutting, cracking, and raveling [26].
However, Pierce and Weitzel [26] implied that further research is required to fully validate
the accuracy and the repeatability of the 3D profiling systems to measure pavement rough-
Remote Sens. 2023, 15, 2941 35 of 44

ness. Additionally, it is necessary to develop a standardized method for calculating IRI

with 3D profiling systems [26]. Thus, this technology is expected to gain more attention to
achieve higher maturity.

5.2.6. Standardization of Equipment Testing

The reported accuracy of the different techniques may not help compare their perfor-
mance due to the absence of standardization in the evaluation process. Many variables
may render the obtained results unrepresentable or make the comparison unfair. First, the
reference roughness measures were obtained differently using rod and level [67,68], laser
profilers [60,61,130], World Bank Class 3 Roughmeter [98], and World Bank Class 3 bump in-
tegrator [84]. Second, there was variability in the roughness reporting measures. Although
the majority of the studies reported roughness data using IRI [41,67,84,112,114,130,192],
some studies reported roughness data using other measures, including PSD [119,126,193],
and RIF [60,100,116]. Third, the scale of the validation case studies varied significantly. For
example, Kumar et al. [57] used data collected from a case study of about 100 km, whereas
Alhasan et al. [125] used two pavement sections of 105 m in total. Thus, it is essential to
standardize the testing of new roughness measurement approaches.

6. Conclusions
Pavement roughness serves as a crucial metric to evaluate the condition of paved
roads and is widely regarded as a primary measure of rideability and serviceability. As a
result, roughness data is in high demand across different industries, including highway
agencies and the automotive industry. Therefore, it is critical to develop techniques that
fulfill the requirement of efficiency, accuracy, and cost-effectiveness that align with the
various applications. Subsequently, the evolution of pavement roughness measurement
techniques has undergone significant changes over the years. The current study utilized
the Scopus database and the TRB library to investigate trends in roughness measurement
techniques among both industry practice and research community over the past five
decades. It also identified the state of the art and the state of the practice in roughness
measurement techniques. Moreover, the study identified current research gaps and industry
needs and anticipated future directions. The study primarily focused on the evolution of
roughness measurement techniques in North America. However, the study supplemented
the discussion by analyzing the worldwide state of the practice via surveying thirty-four
pieces of equipment from eleven countries, including the USA, Canada, Denmark, Germany,
China, Japan, and Australia.
In industry, the use of response-type methods gradually declined over time and
was largely phased out in the 1990s. Non-contact profilers started gaining popularity
in the 1980s and eventually prevailed in the 1990s and 2000s. While ultrasonic sensors
were widely used in non-contact inertial profilers in the 1980s, their usage gradually
decreased, and they were completely replaced by laser sensors from major vendors during
the first decade of this century. Today, laser-based equipment, including laser point and 3D
laser imaging systems, is the most commonly used technology for pavement roughness
measurement. Three-dimensional laser imaging systems based on laser triangulation
technology are increasingly used in commercial systems developed by major vendors due
to their advantages in acquiring multi-use data with high accuracy and efficiency. However,
the anticipated spread of highly autonomous cars and the need for cost-effective techniques
for roughness measurement is expected to increase the potential of using smartphones
and built-in sensors in connected cars, especially in combination with crowdsourcing and
machine learning.
The research community has been increasingly putting significant efforts into inves-
tigating various roughness measurement techniques, particularly in the last few years.
Most research efforts were devoted to evaluating, calibrating, and reviewing response-type
equipment in the 1970s, 1980s and 1990s. However, an increasing number of technologies
have been investigated in recent years. Most of the published studies explored the use
Remote Sens. 2023, 15, 2941 36 of 44

of acceleration signals describing the suspension response of vehicles obtained via smart-
phones, mounted sensors, and vehicles’ built-in sensors. Other studies investigated a wide
range of technologies, including satellite and airborne SAR images, infrared, acoustics,
3D imaging systems, pressure sensors, and pavement-embedded sensors. Developing
low-cost techniques was the focus of multiple studies. Crowdsourcing was also explored
as an alternative approach to using expensive equipment and experienced technicians for
data collection and calibration. Additionally, the study indicates that more research has
been directed toward benefiting parties other than highway agencies. Pavement roughness
evaluation is becoming increasingly important to the automotive industry and navigation.
Although an increasing number of techniques have been investigated in academia,
their robustness in real applications remains dubious due to the lack of standardization
in the evaluation process. Therefore, it is necessary to standardize the evaluation of the
developed technologies. This review indicated little collaboration between industry and
academia in developing new techniques for measuring pavement roughness. Thus, it is
critical to increase cooperation between industry and academia to maximize the benefit
of research efforts in academia. Partnerships can include joint funding, resource sharing,
and co-authoring publications. Government funding agencies can provide incentives for
collaborations by offering funding opportunities and prioritizing joint proposals.

Author Contributions: A.F.: conceptualization, methodology, data collection, analysis, and writing;
T.Z.: review, editing, and supervision. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Not applicable.
Acknowledgments: Ali Fares would like to express his acknowledgment and thanks to the Research
Grants Council (RGC) of Hong Kong for supporting his Ph.D. study through the Hong Kong Ph.D.
Fellowship Scheme (HKPFS).
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A
Table A1. Keyword synonyms.

Label Replace With

Accelerometer Accelerometers
Neural network, neural networks Artificial neural network
Bus lane Bus lanes
Connected vehicle Connected vehicles
Crowdsourced data Crowdsourcing
Dynamic tire pressure sensor (DTPS) Dynamic tire pressure
Frequency response function (FRF) Frequency response function
Global positioning system (GPS), GPS Global positioning system
Laser scanning profilometer Inertial laser profiler
International Roughness Index (IRI), IRI International roughness index
Inverse problem Inverse problems
Kalman filter Kalman filtering
Laser sensor Laser sensors
Longitudinal profile of road, longitudinal road profile Longitudinal profile
Low-cost device Low cost
Remote Sens. 2023, 15, 2941 37 of 44

Table A1. Cont.

Label Replace With

Kinect, Kinect V2, Microsoft Kinect One (V2) Microsoft Kinect
Pavement evaluation, road condition evaluation, road disturbance estimation, road anomaly
Pavement condition assessment
detection
Road health monitoring, road surface monitoring Pavement monitoring
Pavement profiler, profiler, profilers, profile, road profile Pavement profilers
Pavement smoothness, road roughness, road roughness identification, road surface, road surface
Pavement roughness
roughness, roughness, smoothness
Pothole detection Potholes
Power spectrum Power spectral density
Probe vehicle Probe vehicles
Response-type road roughness measuring system (RTRRMS) Response-type device
Riding quality Ride quality
Keywords road profiler Road profiler
Road profile measurement, Roughness assessment, Road Roughness Detection Road roughness estimation
Smartroadsense Smart road sense
Smartphone sensor Smartphone sensors
Smartphone Smartphones
Unmanned aerial vehicles (UAVs) Unmanned aerial vehicles
Vehicle response Vehicle responses
Youla-Kucera (YK) parametrization Youla–Kucera parametrization

References
1. Attoh-Okine, N.; Adarkwa, O. Pavement Condition Surveys-Overview of Current Practices, Delaware Cent; Delaware Center for
Transportation, University of Delaware: Newark, DE, USA, 2013.
2. Pierce, L.M.; Stolte, S.E. NCHRP Synthesis 589: Automated Data Collection and Quality Management for Pavement Condition Reporting;
Transportation Research Board: Washington, DC, USA, 2022. [CrossRef]
3. Mazari, M.; Rodriguez, D.D. Prediction of pavement roughness using a hybrid gene expression programming-neural network
technique. J. Traffic Transp. Eng. Engl. Ed. 2016, 3, 448–455. [CrossRef]
4. Byrne, M.; Albrecht, D.; Sanjayan, J. Identifying error and maintenance intervention of pavement roughness time series with
minimum message length inference. Int. J. Pavement Eng. 2010, 11, 37–47. [CrossRef]
5. New Hampshire Department of Transportation. Transportation Asset Management Plan. 2022. Available online:
https://www.tam-portal.com/collections/tamps/ (accessed on 10 February 2023).
6. Idaho Department of Transportation. Transportation Asset Management Plan. 2019. Available online: https://www.tam-portal.
com/collections/tamps/ (accessed on 10 February 2023).
7. Oklahoma Department of Transportation. Transportation Asset Management Plan. 2019. Available online: https://www.tam-
portal.com/collections/tamps/ (accessed on 10 February 2023).
8. Sakhaeifar, M.S.; Newcomb, D.; Nobakht, M.; Underwood, S.; Gudipudi, P.P.; Stempihar, J. Selection of Long Lasting Reha-Bilitation
Treatment Using Life Cycle Cost Analysis and Present Serviceability Rating; Final Report: FHWA-Ok-15-06; Oklahoma Department of
Transportation: Oklahoma City, OK, USA, 2015.
9. Hood, R. Optimising the Use of Laser Profilometer Data to Report Rut Depth, Roughness and Surface Texture Review of WA Practice;
014509-1; Main Roads Western Australia: East Perth, Australia, 2019.
10. Robbins, M.M.; Tran, N. Literature Review: The Impact of Pavement Roughness on Vehicel Operating Costs; NCAT Report 15-02;
National Asphalt Pavement Association: Montgomery, AL, USA, 2015.
11. Zaabar, I.; Chatti, K. Estimating Vehicle Operating Costs Caused by Pavement Surface Conditions. Transp. Res. Rec. J. Transp. Res.
Board 2014, 2455, 63–76. [CrossRef]
12. Liu, X.; Al-Qadi, I. Integrated Vehicle–Tire–Pavement Approach for Determining Pavement Structure-Induced Rolling Resistance
under Dynamic Loading. Transp. Res. Rec. J. Transp. Res. Board 2022, 2676, 398–409. [CrossRef]
13. Robbins, M.; Tran, N. Review of Initial Service Life Determination in Life (LCCA) Procedures and in Practice; NCAT Report 18-02;
National Asphalt Pavement Association and State Asphalt Pavement Associations: Montgomery, AL, USA, 2018.
14. Hikmah, N.; Tan, S.; Zahran, E.; Yap, Y.; Taib, H. Statistical Correlation between Road Surface Roughness and Traffic Accidents. In
Proceedings of the 7th Brunei International Conference on Engineering and Technology (BICET 2018), Bandar Seri Begawan,
Brunei, 12–14 November 2018; p. 66. [CrossRef]
Remote Sens. 2023, 15, 2941 38 of 44

15. Elghriany, A.; Yi, P.; Liu, P.; Yu, Q. Investigation of the Effect of Pavement Roughness on Crash Rates for Rigid Pavement. J.
Transp. Saf. Secur. 2015, 8, 164–176. [CrossRef]
16. Lee, J.; Abdel-Aty, M.; Nyame-Baafi, E. Investigating the Effects of Pavement Roughness on Freeway Safety using Data from Five
States. Transp. Res. Rec. J. Transp. Res. Board 2020, 2674, 127–134. [CrossRef]
17. Hegmon, R.R. Definition and measurement of pavement surface roughness. Wear 1979, 57, 127–136. [CrossRef]
18. Wambold, J.C. The measurement and data analysis used to evaluate highway roughness. Wear 1979, 57, 117–125. [CrossRef]
19. Hicks, R.G.; Mahoney, J.P. NCHRP Synthesis of Highway Practice 76: Collection and Use of Pavement Condition Data; Transportation
Research Board, National Research Council: Washington, DC, USA, 1981.
20. Campillo, J.R. A Simplified Pavement Condition Assessment and Its Integration to a Pavement Management System. Ph.D.
Dissertation, Arizona State University, Tempe, AZ, USA, 2018.
21. ISO 8608:2016; Mechanical Vibration—Road Surface Profiles—Reporting of Measured Data. International Organization for
Standardization: Genève, Switzerland, 2016. Available online: https://www.iso.org/standard/71202.html (accessed on 20
February 2023).
22. Sayers, M.W. Two Quarter-Car Models for Defining Road Roughness: IRI and HRI. Transp. Res. Rec. 1989, 1215, 165–172.
Available online: https://onlinepubs.trb.org/Onlinepubs/trr/1989/1215/1215-018.pdf (accessed on 20 January 2023).
23. McGhee, K.H. CHRP Synthesis 334: Automated Pavement Distress Collection Techniques; Transportation Research Board: Washington,
DC, USA, 2004. [CrossRef]
24. Fiorentini, N.; Maboudi, M.; Leandri, P.; Losa, M. Can Machine Learning and PS-InSAR Reliably Stand in for Road Profilometric
Surveys? Sensors 2021, 21, 3377. [CrossRef]
25. Epps, J.A.; Monismith, C.L. NCHRP Synthesis of Highway Practice 126: Equipment for Obtaining Pavement Condition and Traffic
Loading Data; Transportation Research Board, National Research Council: Washington, DC, USA, 1986.
26. Pierce, L.M.; Weitzel, N.D. NCHRP Synthesis 531: Automated Pavement Condition Surveys; Transportation Research Board:
Washington, DC, USA, 2019. [CrossRef]
27. Gramling, W.L. NCHRP Synthesis of Highway Practice 203: Current Practices in Determining Pavement Condition, Transportation
Research Board, Washington, DC, 1994. Available online: http://onlinepubs.trb.org/Onlinepubs/nchrp/nchrp_syn_203.pdf
(accessed on 15 February 2023).
28. Nguyen, T.; Lechner, B.; Wong, Y.D. Response-based methods to measure road surface irregularity: A state-of-the-art review. Eur.
Transp. Res. Rev. 2019, 11, 43. [CrossRef]
29. Hettiarachchi, C.; Yuan, J.; Amirkhanian, S.; Xiao, F. Measurement of pavement unevenness and evaluation through the IRI
parameter—An overview. Measurement 2023, 206, 112284. [CrossRef]
30. Yu, Q.; Fang, Y.; Wix, R. Pavement roughness index estimation and anomaly detection using smartphones. Autom. Constr. 2022,
141, 104409. [CrossRef]
31. Echchakoui, S. Why and how to merge Scopus and Web of Science during bibliometric analysis: The case of sales force literature
from 1912 to 2019. J. Mark. Anal. 2020, 8, 165–184. [CrossRef]
32. Elsevier, Scopus, Scopus Content Coverage Guide, Elsevier, B.V., Amesterdam. 2020; Available online: https://www.elsevier.
com/__data/assets/pdf_file/0007/69451/Scopus_ContentCoverage_Guide_WEB.pdf (accessed on 22 March 2021).
33. The National Academies of Sciences, Engineering, and Medicine. The TRB Library. 2023. Available online: https://www.trb.org/
InformationServices/TRBLibrary1.aspx (accessed on 20 February 2023).
34. The National Academies of Sciences, Engineering, and Medicine. TRB Store. 2023. Available online: https://www.mytrb.org/
MyTRB/Store/default.aspx (accessed on 20 March 2023).
35. The National Academies of Sciences, Engineering, and Medicine. NCHRP at 50 Years; Transportation Research Board: Wash-
ington, DC, USA, 2012; Available online: https://onlinepubs.trb.org/onlinepubs/nchrp/docs/NCHRP_50thAnniversary.pdf
(accessed on 5 March 2023).
36. Chang, W.; Cheng, J.; Allaire, J.; Sievert, C.; Schloerke, B.; Xie, Y.; Allen, J.; McPherson, J.; Dipert, A.; Borges, B. Shiny:
Web Application Framework for R. R Package Version 1.7.4.9002. 2023. Available online: https://shiny.rstudio.com/
(accessed on 20 March 2023).
37. Flintsch, G.W.; McGhee, K.K. NCHRP Synthesis 401 Quality: Quality Management of Pavement Condition Data Collection; The National
Academies Press: Washington, DC, USA, 2009. [CrossRef]
38. Federal Highway Administration. National Performance Management Measures; Assessing Pavement Condition for the National
Highway Performance Program and Bridge Condition for the National Highway Performance Program. Fed. Regist. 2017, 82,
5886–5970.
39. González, A.; O’Brien, E.J.; Li, Y.-Y.; Cashell, K. The use of vehicle acceleration measurements to estimate road roughness. Veh.
Syst. Dyn. 2008, 46, 483–499. [CrossRef]
40. Du, Y.; Liu, C.; Wu, D.; Jiang, S. Measurement of International Roughness Index by Using Z-Axis Accelerometers and GPS. Math.
Probl. Eng. 2014, 2014, 928980. [CrossRef]
41. Zang, K.; Shen, J.; Huang, H.; Wan, M.; Shi, J. Assessing and Mapping of Road Surface Roughness based on GPS and Accelerometer
Sensors on Bicycle-Mounted Smartphones. Sensors 2018, 18, 914. [CrossRef] [PubMed]
Remote Sens. 2023, 15, 2941 39 of 44

42. Huang, M.-F.; Ji, P.N.; Wang, T.; Aono, Y.; Salemi, M.; Chen, Y.; Zhao, J.; Xia, T.J.; Wellbrock, G.A.; Huang, Y.-K.; et al. First
Field Trial of Distributed Fiber Optical Sensing and High-Speed Communication Over an Operational Telecom Network. J. Light
Technol. 2019, 38, 75–81. [CrossRef]
43. Harris, N.; Gonzalez, A.; Obrien, E.; McGetrick, P. Characterisation of pavement profile heights using accelerometer readings and
a combinatorial optimisation technique. J. Sound Vib. 2010, 329, 497–508. [CrossRef]
44. Tudon-Martinez, J.C.; Fergani, S.; Sename, O.; Martinez, J.J.; Morales-Menendez, R.; Dugard, L. Adaptive Road Profile Estimation
in Semiactive Car Suspensions. IEEE Trans. Control Syst. Technol. 2015, 23, 2293–2305. [CrossRef]
45. Liu, C.; Wu, D.; Li, Y.; Du, Y. Large-scale pavement roughness measurements with vehicle crowdsourced data using semi-
supervised learning. Transp. Res. Part C Emerg. Technol. 2021, 125, 103048. [CrossRef]
46. Highway Performance Monitoring System Field Manual for the Continuing Analytical and Statistical Database, Appendix J. 1987.
47. Choubane, B.; Scott, S.; Mraz, A.; Schiffermuller, J. Pavement Smoothness Acceptance Testing: Survey of Current State Practices; State
of Florida Department of Transportation: Chipley, FL, USA, 2010.
48. Flintsch, G.W. Mechanistic-Empirical Pavement Design and Performance Prediction, Opportunities and Challenges for PMS. In
Proceedings of the 3rd European Pavement and Asset Management Conference, Coimbra, Portugal, 7–9 July 2008.
49. Federal Highway Administration. HPMS Reassessment 2010+; Final Report; Federal Highway Administration: Washington, DC,
USA, 2008.
50. Woodstrom, J.H. NCHRP Synthesis 167: Measurements, Specifications, and Achievement of Smoothness for Pavement Construction;
Transportation Research Board: Washington, DC, USA, 1990.
51. Mirtabar, Z.; Golroo, A.; Mahmoudzadeh, A.; Barazandeh, F. Development of a crowdsourcing-based system for computing the
international roughness index. Int. J. Pavement Eng. 2020, 23, 489–498. [CrossRef]
52. Bennett, C.R.; De Solminihac, H.; Chamorro, A. Data Collection Technologies for Road Management; World Bank: Washington, DC,
USA, 2007.
53. Ames Engineering. Proven Measurement Products: Sensor Technology, Pavement Measurement. 2023. Available online:
https://amesengineering.com/ (accessed on 25 February 2023).
54. Mahlberg, J.A.; Li, H.; Zachrisson, B.; Leslie, D.K.; Bullock, D.M. Pavement Quality Evaluation Using Connected Vehicle Data.
Sensors 2022, 22, 9109. [CrossRef]
55. Alatoom, Y.I.; Obaidat, T.I. Measurement of Street Pavement Roughness in Urban Areas Using Smartphone. Int. J. Pavement Res.
Technol. 2022, 15, 1003–1020. [CrossRef]
56. Yang, X.; Hu, L.; Ahmed, H.U.; Bridgelall, R.; Huang, Y. Calibration of smartphone sensors to evaluate the ride quality of paved
and unpaved roads. Int. J. Pavement Eng. 2020, 23, 1529–1539. [CrossRef]
57. Kumar, R.; Mukherjee, A.; Singh, V.P. Community Sensor Network for Monitoring Road Roughness Using Smartphones. J.
Comput. Civ. Eng. 2017, 31, 624. [CrossRef]
58. Katicha, S.W.; Khoury, J.; Flintsch, G.W. Assessing the effectiveness of probe vehicle acceleration measurements in estimating
road roughness. Int. J. Pavement Eng. 2015, 17, 698–708. [CrossRef]
59. Fauriat, W.; Mattrand, C.; Gayton, N.; Beakou, A.; Cembrzynski, T. Estimation of road profile variability from measured vehicle
responses. Veh. Syst. Dyn. 2016, 54, 585–605. [CrossRef]
60. Bridgelall, R. Connected Vehicle Approach for Pavement Roughness Evaluation. J. Infrastruct. Syst. 2014, 20, 167. [CrossRef]
61. Chou, C.-P.; Siao, G.-J.; Chen, A.-C.; Lee, C.-C. Algorithm for Estimating International Roughness Index by Response-Based
Measuring Device. J. Transp. Eng. Part B Pavements 2020, 146, 4020031. [CrossRef]
62. Spangler, E.B.; Kelly, W.J. GMR Road Profilometer: A Method for Measuring Road Profile. Highw. Res. Rec. 1964, 121, 27–54.
63. Huft, D.L.; Corcoran, D.C.; Lunde, B.A.; Orth, P.A. Status of the South Dakota Profilometer. Transp. Res. Rec. 1987, 1117, 104–113.
64. ASTM E 950/950M-09; Standard Test Method for Measuring Longitudinal Profile of Traveled Surfaces with an Accelerometer
Established Inertial Profiling Reference. ASTM International: West Conshohocken, PA, USA, 2018. [CrossRef]
65. ASTM E950/E950M-22; Standard Test Method for Measuring the Longitudinal Profile of Traveled Surfaces. ASTM International:
West Conshohocken, PA, USA, 2022. [CrossRef]
66. Mitchell; Chang, J.-R.; Chang, K.-T.; Chen, D.-H. Application of 3D Laser Scanning on Measuring Pavement Roughness. J. Test.
Eval. 2006, 34, jte13178. [CrossRef]
67. Khalifeh, V.; Golroo, A.; Ovaici, K. Application of an Inexpensive Sensor in Calculating the International Roughness Index. J.
Comput. Civ. Eng. 2018, 32, 761. [CrossRef]
68. Mahmoudzadeh, A.; Golroo, A.; Jahanshahi, M.R.; Yeganeh, S.F. Estimating Pavement Roughness by Fusing Color and Depth
Data Obtained from an Inexpensive RGB-D Sensor. Sensors 2019, 19, 1655. [CrossRef]
69. Aleadelat, W.; Aledealat, K.; Ksaibati, K. Estimating pavement roughness using a low-cost depth camera. Int. J. Pavement Eng.
2022, 23, 4923–4930. [CrossRef]
70. Guo, R.; Yu, Z.; Zhou, Y. Development and Preliminary Evaluation of a Varying-Speed Road Profiler. J. Test. Eval. 2020, 48,
3479–3489. [CrossRef]
71. Prosser-Contreras, M.; Atencio, E.; La Rivera, F.M.; Herrera, R.F. Use of Unmanned Aerial Vehicles (UAVs) and Photogrammetry
to Obtain the International Roughness Index (IRI) on Roads. Appl. Sci. 2020, 10, 8788. [CrossRef]
72. Ma, N.; Fan, J.; Wang, W.; Wu, J.; Jiang, Y.; Xie, L.; Fan, R. Computer vision for road imaging and pothole detection: A
state-of-the-art review of systems and algorithms. Transp. Saf. Environ. 2022, 4, tdac026. [CrossRef]
Remote Sens. 2023, 15, 2941 40 of 44

73. Huft, D.L. South Dakota Profilometer. Transp. Res. Rec. 1984, 1–8. Available online: https://onlinepubs.trb.org/Onlinepubs/trr/
1984/1000/1000-001.pdf (accessed on 12 February 2023).
74. Bloom, J.A.; Schwartz, P.C. Development of a Noncontact Pavement Smoothness Monitor for Use During Construction. Transp.
Res. Rec. 1984, 986, 18–22.
75. Brown, D. Evaluation of the PRORUT System. Public Roads 1990, 53, 118–122.
76. Walker, R.S.; Lin, H.-T. Profilograph Correlation Study with Present Serviceability Index. Transp. Res. Rec. 1988, 1196, 257–275.
Available online: http://onlinepubs.trb.org/Onlinepubs/trr/1988/1196/1196-026.pdf (accessed on 12 February 2023).
77. Moore, R.K.; Clark, G.N.; Plumb, G.N. Present Serviceability-Roughness Correlations Using Rating Panel Data. Transp. Res. Rec.
1987, 1117, 152–158.
78. Spangler, E.B.; Rizenbergs, R.L.; Burchett, J.L.; Robinson, D.C. Use of the Inertial Profilometer to Calibrate Kentucky Department of
Highways Mays Ride Meter Systems. Transp. Res. Rec. 1988, 286–293. Available online: https://onlinepubs.trb.org/Onlinepubs/
trr/1988/1196/1196-028.pdf (accessed on 12 February 2023).
79. Kulakowski, B.T.; Henry, J.J.; Wambold, J.C. Relative Influence of Accelerometer and Displacement Transducer Signals in Road
Roughness Measurements. Transp. Res. Rec. 1988, 1196, 313–317. Available online: http://onlinepubs.trb.org/Onlinepubs/trr/19
88/1196/1196-032.pdf (accessed on 12 February 2023).
80. Lenngren, C.A. Some Experience Using Noncontact Sensors on Airport Pavements. In Proceedings of the Nondestructive
Evaluation of Aging Aircraft, Airports, and Aerospace Hardware, Scottsdale, AZ, USA, 3–5 December 1996; pp. 250–260.
[CrossRef]
81. Šprinc, J.; Kropác, O.; Šprinc, M. Characterization of Longitudinal Road Unevenness in the Light of the International PIARC?
EVEN Experiment 1998. Veh. Syst. Dyn. 2002, 37, 263–281. [CrossRef]
82. Ahlin, K.; Granlund, J.; Lindstrom, F. Comparing road profiles with vehicle perceived roughness. Int. J. Veh. Des. 2004, 36, 270.
[CrossRef]
83. Mitchell, D.; Chen, J.-S.; Chang, M.-K.; Lin, K.-Y. Device Harmonization to Determine Smoothness Specification for Asphalt
Pavements. J. Test. Eval. 2006, 34, jte100031. [CrossRef]
84. Sandamal, R.M.K.; Pasindu, H.R. Applicability of smartphone-based roughness data for rural road pavement condition evaluation.
Int. J. Pavement Eng. 2020, 23, 663–672. [CrossRef]
85. Daraghmi, Y.-A.; Wu, T.-H.; Ik, T.-U. Crowdsourcing-Based Road Surface Evaluation and Indexing. IEEE Trans. Intell. Transp. Syst.
2020, 23, 4164–4175. [CrossRef]
86. Alatoom, Y.I.; Obaidat, T.I.A.-S. Development of pavement roughness models using Artificial Neural Network (ANN). Int. J.
Pavement Eng. 2022, 23, 4622–4637. [CrossRef]
87. Grabowski, D.; Szczodrak, M.; Czyzewski, A. Economical Methods for Measuring Road Surface Roughness. Metrol. Meas. Syst.
2018, 25, 533–549. [CrossRef]
88. Bridgelall, R.; Hough, J.; Tolliver, D. Characterising pavement roughness at non-uniform speeds using connected vehicles. Int. J.
Pavement Eng. 2019, 20, 958–964. [CrossRef]
89. Zhao, B.; Nagayama, T.; Toyoda, M.; Makihata, N.; Takahashi, M.; Ieiri, M. Vehicle Model Calibration in the Frequency Domain
and its Application to Large-Scale IRI Estimation. J. Disaster Res. 2017, 12, 446–455. [CrossRef]
90. Medina, J.R.; Salim, R.; Underwood, B.S.; Kaloush, K. Experimental Study for Crowdsourced Ride Quality Index Estimation
Using Smartphones. J. Transp. Eng. Part B Pavements 2020, 146, 04020070. [CrossRef]
91. Medina, J.R.; Noorvand, H.; Underwood, B.S.; Kaloush, K. Statistical Validation of Crowdsourced Pavement Ride Quality
Measurements from Smartphones. J. Comput. Civ. Eng. 2020, 34, 891. [CrossRef]
92. Zhang, Z.; Zhang, H.; Xu, S.; Lv, W. Pavement roughness evaluation method based on the theoretical relationship between
acceleration measured by smartphone and IRI. Int. J. Pavement Eng. 2022, 23, 3082–3098. [CrossRef]
93. Ahmed, H.U.; Hu, L.; Yang, X.; Bridgelall, R.; Huang, Y. Effects of smartphone sensor variability in road roughness evaluation.
Int. J. Pavement Eng. 2022, 23, 4404–4409. [CrossRef]
94. Hanson, T.; Cameron, C.; Hildebrand, E. Evaluation of low-cost consumer-level mobile phone technology for measuring
international roughness index (IRI) values. Can. J. Civ. Eng. 2014, 41, 819–827. [CrossRef]
95. Bisconsini, D.R.; Pegorini, V.; Casanova, D.; de Oliveira, R.A.; Farias, B.A.; Júnior, J.L.F. Intervening Factors in Pavement
Roughness Assessment with Smartphones: Quantifying the Effects and Proposing Mitigation. J. Transp. Eng. Part B Pavements
2021, 147, 04021051. [CrossRef]
96. Zeng, H.; Park, H.; Smith, B.L. Impact of Vehicle Dynamic Systems on a Connected Vehicle-Enabled Pavement Roughness
Estimation. J. Infrastruct. Syst. 2019, 25, 461. [CrossRef]
97. Botshekan, M.; Asaadi, E.; Roxon, J.; Ulm, F.-J.; Tootkaboni, M.; Louhghalam, A. Smartphone-enabled road condition monitoring:
From accelerations to road roughness and excess energy dissipation. Proc. R. Soc. A Math. Phys. Eng. Sci. 2021, 477, 20200701.
[CrossRef]
98. Janani, L.; Doley, R.; Sunitha, V.; Mathew, S. Precision enhancement of smartphone sensor-based pavement roughness estimation
by standardizing host vehicle speed. Can. J. Civ. Eng. 2022, 49, 716–730. [CrossRef]
99. Alessandroni, G.; Carini, A.; Lattanzi, E.; Freschi, V.; Bogliolo, A. A Study on the Influence of Speed on Road Roughness Sensing:
The SmartRoadSense Case. Sensors 2017, 17, 305. [CrossRef]
100. Bridgelall, R. Inertial Sensor Sample Rate Selection for Ride Quality Measures. J. Infrastruct. Syst. 2015, 21, 225. [CrossRef]
Remote Sens. 2023, 15, 2941 41 of 44

101. Zeng, H.; Park, H.; Smith, B.L.; Parkany, E. Feasibility Assessment of a Smartphone-Based Application to Estimate Road
Roughness. KSCE J. Civ. Eng. 2018, 22, 3120–3129. [CrossRef]
102. Jeong, J.; Jo, H.; Ditzler, G. Convolutional neural networks for pavement roughness assessment using calibration-free vehicle
dynamics. Comput. Civ. Infrastruct. Eng. 2020, 35, 1209–1229. [CrossRef]
103. Opara, K.R.; Brzezinski, K.; Bukowicki, M.; Kaczmarek-Majer, K. Road Roughness Estimation Through Smartphone-Measured
Acceleration. IEEE Intell. Transp. Syst. Mag. 2022, 14, 209–220. [CrossRef]
104. Gamage, Y.T.; Thotawaththa, T.A.I.; Wijayasiri, A. Measuring Road Roughness through Crowdsourcing while Minimizing the
Conditional Effects. Int. J. Intell. Transp. Syst. Res. 2022, 20, 581–601. [CrossRef]
105. Liu, Q.; Sun, L.; Kornhauser, A.; Sun, J.; Sangwa, N. Road roughness acquisition and classification using improved restricted
Boltzmann machine deep learning algorithm. Sens. Rev. 2019, 39, 733–742. [CrossRef]
106. Nguyen, T.; Lechner, B.; Wong, Y.D.; Tan, J.Y. Bus Ride Index—A refined approach to evaluating road surface irregularities. Road
Mater. Pavement Des. 2021, 22, 423–443. [CrossRef]
107. Chen, S.; Xue, J. Road Roughness Level Identification Based on BiGRU Network. IEEE Access 2022, 10, 32696–32705. [CrossRef]
108. Liu, C.; Wu, D.; Li, Y.; Jiang, S.; Du, Y. Mathematical insights into the relationship between pavement roughness and vehicle
vibration. Int. J. Pavement Eng. 2022, 23, 1935–1947. [CrossRef]
109. Bidgoli, M.A.; Golroo, A.; Nadjar, H.S.; Rashidabad, A.G.; Ganji, M.R. Road roughness measurement using a cost-effective
sensor-based monitoring system. Autom. Constr. 2019, 104, 140–152. [CrossRef]
110. Zhang, Q.; Hou, J.; Hu, X.; Yuan, L.; Jankowski, Ł.; An, X.; Duan, Z. Vehicle parameter identification and road roughness
estimation using vehicle responses measured in field tests. Measurement 2022, 199, 111348. [CrossRef]
111. Li, J.; Zhang, Z.; Wang, W. New Approach for Estimating International Roughness Index Based on the Inverse Pseudo Excitation
Method. J. Transp. Eng. Part B Pavements 2019, 145, 04018059. [CrossRef]
112. Du, Y.; Liu, C.; Wu, D.; Li, S. Application of Vehicle Mounted Accelerometers to Measure Pavement Roughness. Int. J. Distrib.
Sens. Networks 2016, 12, 8413146. [CrossRef]
113. Liu, W.; Wang, R.; Ding, R.; Meng, X.; Yang, L. On-line estimation of road profile in semi-active suspension based on unsprung
mass acceleration. Mech. Syst. Signal Process. 2020, 135, 106370. [CrossRef]
114. Zhang, Z.; Sun, C.; Bridgelall, R.; Sun, M. Application of a Machine Learning Method to Evaluate Road Roughness from
Connected Vehicles. J. Transp. Eng. Part B Pavements 2018, 144, 04018043. [CrossRef]
115. Nitsche, P.; Stütz, R.; Kammer, M.; Maurer, P. Comparison of Machine Learning Methods for Evaluating Pavement Roughness
Based on Vehicle Response. J. Comput. Civ. Eng. 2014, 28, 285. [CrossRef]
116. Bridgelall, R.; Rahman, T.; Tolliver, D.; Daleiden, J.F. Wavelength sensitivity of roughness measurements using connected vehicles.
Int. J. Pavement Eng. 2019, 20, 566–572. [CrossRef]
117. Wang, G.; Burrow, M.; Ghataora, G. Study of the Factors Affecting Road Roughness Measurement Using Smartphones. J.
Infrastruct. Syst. 2020, 26, 04020020. [CrossRef]
118. Li, Z.; Yu, W.; Cui, X. Online Classification of Road Roughness Conditions with Vehicle Unsprung Mass Acceleration by Sliding
Time Window. Shock. Vib. 2018, 2018, 5131434. [CrossRef]
119. Yousefzadeh, M.; Azadi, S.; Soltani, A. Road profile estimation using neural network algorithm. J. Mech. Sci. Technol. 2010, 24,
743–754. [CrossRef]
120. Wei, W.; Shaoyi, B.; Lanchun, Z.; Yongzhi, W.; Hui, Y. Pavement Roughness Identification Research in Time Domain Based on
Neural Network. J. Vibroeng. 2015, 17, 3865–3875.
121. Zennaro, S.; Munaro, M.; Milani, S.; Zanuttigh, P.; Bernardi, A.; Ghidoni, S.; Menegatti, E. Performance evaluation of the 1st and
2nd generation Kinect for multimedia applications. In Proceedings of the 2015 IEEE International Conference on Multimedia and
Expo (ICME), Turin, Italy, 29 June–3 July 2015; pp. 1–6.
122. Tran, T.H.; Taweep, C. Automated extraction of expressway road surface from mobile laser scanning data. J. Cent. South Univ.
2020, 27, 1917–1938. [CrossRef]
123. Kumar, P.; Lewis, P.; McElhinney, C.P.; Rahman, A.A. An Algorithm for Automated Estimation of Road Roughness from Mobile
Laser Scanning Data. Photogramm. Rec. 2015, 30, 30–45. [CrossRef]
124. Cahalane, C.; McCarthy, T.; Mc Elhinney, C.P. Mobile mapping system performance: An initial investigation into the effect of
vehicle speed on laser scan lines. In Proceedings of the Remote Sensing & Photogrammety Society Annual Conference—From the
Sea-Bed to the Cloudtops, Cork, Ireland, 1–3 September 2010.
125. Alhasan, A.; White, D.J.; De Brabanter, K.; Alhsan, A. Spatial pavement roughness from stationary laser scanning. Int. J. Pavement
Eng. 2017, 18, 1065403. [CrossRef]
126. Lee, J.-H.; Lee, S.-H.; Kang, D.-K.; Na, S.-D.; Yoo, W.-S. Development of a 3D road profile measuring system for unpaved road
severity analysis. Int. J. Precis. Eng. Manuf. 2017, 18, 155–162. [CrossRef]
127. Abohamer, H.; Elseifi, M.; Dhakal, N.; Zhang, Z.; Fillastre, C.N. Development of a Deep Convolutional Neural Network for the
Prediction of Pavement Roughness from 3D Images. J. Transp. Eng. Part B Pavements 2021, 147, 04021048. [CrossRef]
128. Botha, T.R.; Els, P.S. Rough terrain profiling using digital image correlation. J. Terramechanics 2015, 59, 1–17. [CrossRef]
129. Meyer, F.J.; Ajadi, O.A.; Hoppe, E.J. Studying the Applicability of X-Band SAR Data to the Network-Scale Mapping of Pavement
Roughness on US Roads. Remote Sens. 2020, 12, 1507. [CrossRef]
Remote Sens. 2023, 15, 2941 42 of 44

130. Bashar, M.Z.; Torres-Machi, C. Deep learning for estimating pavement roughness using synthetic aperture radar data. Autom.
Constr. 2022, 142, 104504. [CrossRef]
131. Suanpaga, W.; Yoshikazu, K. Riding Quality Model for Asphalt Pavement Monitoring Using Phase Array Type L-band Synthetic
Aperture Radar (PALSAR). Remote Sens. 2010, 2, 2531–2546. [CrossRef]
132. Karimzadeh, S.; Matsuoka, M. Development of Nationwide Road Quality Map: Remote Sensing Meets Field Sensing. Sensors
2021, 21, 2251. [CrossRef]
133. Babu, A.; Baumgartner, S.V.; Krieger, G. Approaches for Road Surface Roughness Estimation Using Airborne Polarimetric SAR.
IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2022, 15, 3444–3462. [CrossRef]
134. Karimzadeh, S.; Matsuoka, M. Remote Sensing X-Band SAR Data for Land Subsidence and Pavement Monitoring. Sensors 2020,
20, 4751. [CrossRef]
135. Rischioni, L.G.; Babu, A.; Baumgartner, S.V.; Krieger, G. Machine Learning Approaches for Road Condition Monitoring Using
Synthetic Aperture Radar. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 3070–3082. [CrossRef]
136. Zeng, Q.; Hu, X.; Shi, X.; Ren, Y.; Li, Y.; Duan, Z. Estimation of Road Roughness Based on Tire Pressure Monitoring. Int. J. Struct.
Stab. Dyn. 2022, 22, 22500730. [CrossRef]
137. Zhao, Y.; McDaniel, J.G.; Wang, M.L. IRI estimation using analysis of dynamic tire pressure and axle acceleration. Smart Struct.
Syst. 2017, 19, 151–161. [CrossRef]
138. Zhao, Y.; Wang, M.L. IRI measurement using dynamic tire pressure sensor with an axle accelerometer. J. Civ. Struct. Health Monit.
2016, 6, 791–802. [CrossRef]
139. Yang, S.; Wang, R.; Shi, R.; Chen, Y.; Lu, J.; Li, Z.; Cao, Y. An intelligent tyre system for road condition perception. Int. J. Pavement
Eng. 2022, 1–12. [CrossRef]
140. Becker, C.; Els, S. Profiling of rough terrain. Int. J. Veh. Des. 2014, 64, 240. [CrossRef]
141. Zhao, Y.; McDaniel, J.G.; Wang, M.L. IRI Estimation Using Probabilistic Analysis of Acoustic Measurements. Mater. Perform.
Charact. 2013, 2, 339–359. [CrossRef]
142. Wang, H.-P.; Xiang, P.; Jiang, L.-Z. Optical fiber sensing technology for full-scale condition monitoring of pavement layers. Road
Mater. Pavement Des. 2020, 21, 1258–1273. [CrossRef]
143. Zhang, Z.; Deng, F.; Huang, Y.; Bridgelall, R. Road roughness evaluation using in-pavement strain sensors. Smart Mater. Struct.
2015, 24, 115029. [CrossRef]
144. BumpRecorder Co., Ltd. BumpRecorder. 2023. Available online: http://www.bumprecorder.com/en/ (accessed on 8 May 2023).
145. ARRB Systems. Hawkeye 2000. 2023. Available online: https://arrbsystems.com/fact-sheet/hawkeye-2000/ (accessed on 25
February 2023).
146. Mandli Communications. Mandli Collection System. 2023. Available online: https://www.mandli.com/ (accessed on 25 February
2023).
147. International Cybernetics Co. IrisGO Pave. 2023. Available online: https://www.internationalcybernetics.com/iris-gopave/
(accessed on 25 February 2023).
148. International Cybernetics Co. IrisPRO Pave. 2023. Available online: https://www.internationalcybernetics.com/iris-propave/
(accessed on 25 February 2023).
149. WayLink Systems, Co. PVision3D Application. 2023. Available online: http://www.pvision3d.com/ (accessed on 4 April 2023).
150. Gui, R.; Xu, X.; Zhang, D.; Lin, H.; Pu, F.; He, L.; Cao, M. A Component Decomposition Model for 3D Laser Scanning Pavement
Data Based on High-Pass Filtering and Sparse Analysis. Sensors 2018, 18, 2294. [CrossRef]
151. AL-Engineering Oy. Roadmaster Road Roughness Meter. 2023. Available online: http://www.al-engineering.fi/en/roadmaster.
html (accessed on 8 May 2023).
152. ARRB Systems. Roughometer 4. 2023. Available online: https://arrbsystems.com/fact-sheet/roughometer-4/ (accessed on 25
February 2023).
153. ARRB Systems. Walking Profiler G3. 2023. Available online: https://arrbsystems.com/fact-sheet/walking-profiler-g3/
(accessed on 25 February 2023).
154. Zhuozhou Tianpeng Instrument Manufacturing Co., Ltd. DAPPES-3. 2023. Available online: https://www.testmould.com/test_
on_site/walking-profiler.html (accessed on 8 May 2023).
155. International Cybernetics, Co. SurPRO. 2023. Available online: https://www.internationalcybernetics.com/surpro/
(accessed on 25 February 2023).
156. Surface Systems & Instruments Inc. CS8800 Walking Profiler. 2023. Available online: https://www.smoothroad.com/equipment/
walking-profilers/#/tab=products
(accessed on 25 February 2023).
157. Dynatest. Road Surface Profiler MK III. 2023. Available online: https://dynatest.com/equipment/road-surface-profiler-2/
(accessed on 25 February 2023).
158. Dynatest. Road Surface Profiler IV. 2023. Available online: https://dynatest.com/equipment/road-surface-profiler/
(accessed on 25 February 2023).
159. Nanjing T-Bota Scietech Instruments & Equipment Co., Ltd. Road Surface Profiler. 2019. Available online: https://www.
tbtscietech.com/Road-Surface-Profilometer-RSP-pd90700436.html (accessed on 8 May 2023).
160. PaveTesting Ltd. PaveProf V2.0. 2023. Available online: https://pavetesting.com/ (accessed on 25 February 2023).
Remote Sens. 2023, 15, 2941 43 of 44

161. ARRB Systems. Digital Laser Profiler. 2023. Available online: https://arrbsystems.com/fact-sheet/hawkeye-1000/
(accessed on 25 February 2023).
162. Greenwood Engineering. LaserProf. 2023. Available online: https://greenwood.dk/road/laserprof/ (accessed on 25 February
2023).
163. Greenwood Engineering. Greenwood Profiler. 2023. Available online: https://greenwood.dk/road/profiler/ (accessed on 25
February 2023).
164. International Cybernetics Co. IrisGO Portable Profiler. 2023. Available online: https://www.internationalcybernetics.com/irisgo-
portable/ (accessed on 25 February 2023).
165. International Cybernetics, Co. IrisPRO Inertial Profiler. 2023. Available online: https://www.internationalcybernetics.com/
inertial-profilers/ (accessed on 25 February 2023).
166. Cafiso, S.; DI Graziano, A.; D’Agostino, C.; Pappalardo, G.; Delfino, E. A new perspective in the road asset management with the
use of advanced monitoring system & BIM. MATEC Web Conf. 2018, 231, 01007. [CrossRef]
167. Surface Systems & Instruments Inc. CS9550 Portable Survey Profiler. 2023. Available online: https://www.smoothroad.com/
equipment/mobile-surveying/cs9550-portable-survey-profiler/#/tab=overview (accessed on 25 February 2023).
168. Surface Systems & Instruments Inc. CS8600 Ultralight Surface Profiling System. 2023. Available online: https://www.smoothroad.
com/equipment/ada-profilers/cs8600-ultra-light-profiler/ (accessed on 25 February 2023).
169. Data Collection Limited. ROMDAS Laser Profilometer. 2023. Available online: https://romdas.com/romdas-laser-profiler.html
(accessed on 25 February 2023).
170. Applied Pavement Technology Inc. APTech’s EDGE. 2023. Available online: https://www.appliedpavement.com/the-aptech-
edge (accessed on 8 May 2023).
171. Serigos, P.A.; Prozzi, J.A.; Nam, B.H.; Murphy, M.R. Field Evaluation of Automated Rutting Measuring Equipment; No. FHWA/TX-
12/0-6663-1; Texas Department of Transportation: Austin, TX, USA, 2012.
172. Ames Engineering. 8300 High Speed Inertial Road Profiler. 2023. Available online: https://amesengineering.com/ (accessed on
25 February 2023).
173. Ames Engineering. 6300 Lightweight Profiler. 2023. Available online: https://amesengineering.com/products/6300-lightweight-
profiler/ (accessed on 25 February 2023).
174. Pathway Services Inc. PathRunner. 2023. Available online: https://www.pathwayservices.com/technology/ (accessed on 25
February 2023).
175. Fraunhofer IPM. Pavement Profile Scanner PPS. 2023. Available online: https://www.ipm.fraunhofer.de/en.html (accessed on 25
February 2023).
176. LMI Technologies. A Global Leader in 3D Scanning and Inspection. 2023. Available online: https://lmi3d.com/wp-content/
uploads/2016/08/BROCHURE_RoadandRail_US_WEB.pdf (accessed on 25 February 2023).
177. Pavemetrics. Laser Crack Measurement System (LCMS®-2). 2023. Available online: https://www.pavemetrics.com/applications/
road-inspection/lcms2-en/ (accessed on 25 February 2023).
178. Fiorentini, N.; Maboudi, M.; Leandri, P.; Losa, M.; Gerke, M. Surface Motion Prediction and Mapping for Road Infrastructures
Management by PS-InSAR Measurements and Machine Learning Algorithms. Remote Sens. 2020, 12, 3976. [CrossRef]
179. Gao, H.; Qiu, B.; Barroso, R.J.D.; Hussain, W.; Xu, Y.; Wang, X. TSMAE: A Novel Anomaly Detection Approach for Internet of
Things Time Series Data Using Memory-Augmented Autoencoder. IEEE Trans. Netw. Sci. Eng. 2022. [CrossRef]
180. Rathee, G.; Sharma, A.; Iqbal, R.; Aloqaily, M.; Jaglan, N.; Kumar, R. A Blockchain Framework for Securing Connected and
Autonomous Vehicles. Sensors 2019, 19, 3165. [CrossRef] [PubMed]
181. Toulouse, M.; Dai, H.; Le, T.G. Distributed load-balancing for account-based sharded blockchains. Int. J. Web Inf. Syst. 2022, 18,
100–116. [CrossRef]
182. Xu, W.; Duan, H.; Chen, X.; Huang, J.; Liu, D.; Chen, Y. Blockchain-based multi-skill mobile crowdsourcing services. EURASIP J.
Wirel. Commun. Netw. 2022, 2022, 55. [CrossRef]
183. Behera, H.K.; Pradhan, S.; Das, S.S. Low cost ultrasonic roughometer for pavement roughness measurement. Innov. Infrastruct.
Solut. 2021, 6, 168. [CrossRef]
184. Kheirati, A.; Golroo, A. Low-cost infrared-based pavement roughness data acquisition for low volume roads. Autom. Constr. 2020,
119, 103363. [CrossRef]
185. Wessels, I.; Steyn, W.J. Continuous, response-based road roughness measurements utilising data harvested from telematics device
sensors. Int. J. Pavement Eng. 2020, 21, 437–446. [CrossRef]
186. Janani, L.; Sunitha, V.; Mathew, S. Influence of surface distresses on smartphone-based pavement roughness evaluation. Int. J.
Pavement Eng. 2021, 22, 1637–1650. [CrossRef]
187. Genser, A.; Spielhofer, R.; Nitsche, P.; Kouvelas, A. Ride comfort assessment for automated vehicles utilizing a road surface model
and Monte Carlo simulations. Comput. Civ. Infrastruct. Eng. 2022, 37, 1316–1334. [CrossRef]
188. Rahmani, O.; Tehrani, H.G.; Nasiri, A.S.A. A new procedure for analysis of ride quality in roads using multi-body dynamic
simulation. Innov. Infrastruct. Solut. 2022, 7, 209. [CrossRef]
189. Jiang, J.; Seaid, M.; Mohamed, M.S.; Li, H. Inverse algorithm for real-time road roughness estimation for autonomous vehicles.
Arch. Appl. Mech. 2020, 90, 1333–1348. [CrossRef]
Remote Sens. 2023, 15, 2941 44 of 44

190. Theunissen, J.; Tota, A.; Gruber, P.; Dhaens, M.; Sorniotti, A. Preview-based techniques for vehicle suspension control: A
state-of-the-art review. Annu. Rev. Control 2021, 51, 206–235. [CrossRef]
191. Qi, S.; Li, W.; Zhu, A.; Bai, X.F. Responses of Preview-based Vehicle Suspension System on Discrete Impact Roads. In Proceedings
of the 6th CAA International Conference on Vehicular Control and Intelligence (CVCI2022), Nanjing, China, 28–30 October 2022;
IEEE: Piscataway Township, NJ, USA; pp. 1–6. [CrossRef]
192. Jagadeesh Babu, M.; Rahul, B.G. Analysis of Roughness Index of a Flexible Pavement Using Low Cost Instrumentation. Int. J. Civ.
Eng. Technol. 2017, 8, 2036–2043.
193. Liu, Y.; Cui, D. Research on Road Roughness Based on NARX Neural Network. Math. Probl. Eng. 2021, 2021, 9173870. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.
Reproduced with permission of copyright owner. Further reproduction
prohibited without permission.