Hindawi

Journal of Advanced Transportation

Volume 2020, Article ID 1361583, 13 pages
https://doi.org/10.1155/2020/1361583

Research Article
Freeway Traffic Speed Estimation of Mixed Traffic Using
Data from Connected and Autonomous Vehicles with a Low
Penetration Rate

Shanglu He ,1,2 Xiaoyu Guo,3 Fan Ding,4,5 Yong Qi,1 and Tao Chen2
1
School of Automation, Nanjing University of Science and Technology, Nanjing 210094, China
2
Key Laboratory for Automotive Transportation Safety Enhancement Technology of the Ministry of Communication,
Chang’an University, Xi’an 710084, China
3
Zachry Department of Civil and Environmental Engineering, Texas A&M University, College Station, TX 77843-3135, USA
4
Joint Research Institute on Internet of Mobility, Southeast University and University of Wisconsin-Madison, Madison, WI, USA
5
School of Transportation, Southeast University, Nanjing 211189, China

Correspondence should be addressed to Shanglu He; [email protected]

Received 12 December 2019; Revised 20 March 2020; Accepted 25 May 2020; Published 13 June 2020

Academic Editor: Guohui Zhang

Copyright © 2020 Shanglu He et al. *is is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Connected and autonomous vehicles (CAVs) are on the way to the field application. In the beginning stage, there will be a mixed
traffic flow, containing the regular human-driven vehicles and CAVs with a low penetration rate. Recently, the discussion about
the impact of a small proportion of CAVs in the mixed traffic is controversial. *is paper investigated the possibility of applying
the limited data from these lowly penetrated CAVs to estimate the average freeway link speeds based on the Kalman filtering (KF)
method. First, this paper established a VISSIM-based microsimulation model to mimic the mixed traffic with different CAV
penetration rates. *e characteristics of this mixed traffic were then discussed based on the simulation data, including the sample
size distribution, data-missing rate, speed difference, and fundamental diagram. Accordingly, the traditional KF-based method
was introduced and modified to adapt data from CAVs. Finally, the evaluations of the estimation accuracy and the sensitive
analysis of the proposed method were conducted. *e results revealed the possibility and applicability of link speed estimation
using data from a small proportion of CAVs.

1. Introduction there will be a long period of mixed traffic condition

comprising CAVs and regular human-driven vehicles (RVs).
Autonomous vehicle (AV) technology is a hot and practical A majority of research works have been dedicated to the
research spot. When AVs embedded with the feature to impact analysis of AVs/CAVs in the mixed traffic. Some
communicate with others including vehicles, roadside in- focused on the impact on the traffic efficiency, i.e., capacity
frastructures, or traffic control centers, they are defined as and throughput. For instance, Davis explored the contri-
connected and autonomous vehicles (CAVs). It is expected bution of adaptive cruise control (ACC) vehicles to the
that CAVs can provide faster responses and keep shorter reduction in the jam formation [6]. Shladover et al. proved
headways, which lead to an increased overall roadway ca- that the Cooperative Adaptive Cruise Control (CACC)
pacity [1]. Other expected benefits of CAVs include im- technology has the potential to increase lane throughput
proved mobility to people with disabilities, enhanced from the average 2000 veh/h to approximately 4000 veh/h
productive use of travel time, better fuel efficiency, fewer with high market penetrations [7]. Friedrich found that the
emissions, and flexibility in parking [2, 3]. However, it is traffic volume could be increased to about 3900 veh/h/lane
estimated that the market penetration rate of CAVs might be when AVs are in application compared with current
able to reach between 24% and 87% by 2045 [4, 5]. *erefore, designed capacity values of a lane of 2200 veh/h [8]. Both
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 Journal of Advanced Transportation

Zhou et al. and Xiao et al. found that a cooperative control of However, the CAVs have not been put into the market offi-
AVs would improve the traffic efficiency of the merging area cially, so it is hard to achieve sufficient historical CAV data.
[9, 10]. Some focused on modelling the different traffic Regarding these factors, this study would like to focus on the
behaviours of CAVs, such as fundamental diagram and model-based estimation method using CAV data. *ere has
longitudinal and latitudinal movements. For example, been some research using the model-based method. For in-
Baskar et al. demonstrated that RVs and ACC-equipped stance, Wang et al. compared the first-order and second-order
intelligent vehicles had the different fundamental diagrams models to estimate the mixed traffic state with different AV
[11]. Liu et al. changed the lane-changing rules in the cellular penetration rates [25], but they did not discuss the low pen-
automata to simulate the autonomous vehicles [12]. Lu and etration specifically. Considering the controversy under the low
Aakre proposed a smart driver model to simulate the car- penetration condition, this study aims to furtherly discuss how
following behaviour of CAVs [13]. Moreover, some dis- to use the model-based estimation method with information
cussed the influence on other aspects, i.e., safety and en- from a small proportion of CAVs in mixed traffic.
vironment. For example, Morando et al. investigated the More specifically, this study would firstly contribute to set
safety performance of AVs with varying penetration rates in up a simulation platform. Hereafter, this study would explore
two different cases, i.e., a roundabout and a signalized in- the sampling characteristics of CAV probes under a low
tersection [14]. Lu et al. improved the ACC model of CAVs penetration rate, such as their sample size, data-missing rate,
and validated that these CAVs performed better than RVs in and their speed difference with the average link speeds. Fur-
fuel economy [15, 16]. thermore, whether their limited information is supportive to
Obviously, these works admitted that CAVs had different the traffic state estimation would be discussed. Afterwards,
behaviours compared with RVs. Most of them expected the although the KF technique is widely used, this study makes the
CAVs to have a faster reaction time, and thus, CAVs could following adjustments to adapt the lowly penetrated CAVs: a
keep a smaller distance with the front vehicle and be safer. recursive model to fulfil the missing parts, calculation methods
*ese works have validated that the application of CAVs is for state, and measurement noise. Its performance and accu-
definitely beneficial when CAVs take a high proportion of racy are going to be evaluated.
traffic, but the impact of CAVs with a low penetration rate is Accordingly, the rest of this paper is organized as follows:
controversial. If the penetration rate of CAVs is high in the Section 2 introduces a simulation platform of mixed traffic to
mixed traffic, the information from CAVs is definitely suffi- generate the data for the following investigations. *e different
cient to identify the traffic state. What if CAVs only take a low characteristics of traffic with CAVs are discussed in Section 3.
proportion of mixed traffic, will their information be enough to Section 4 presents the exploration of KF-based estimation.
acquire or estimate the traffic state? Since the penetration rate Finally, Section 5 summarizes the main conclusions and
of CAVs grows slowly, it is meaningful to explore whether provides some plans for the improvement and future study.
these CAVs in low penetration rates are a new data source to
assist the surveillance of the traffic condition.
Data provided by CAVs resemble the data collected
2. Simulation Platform for Mixed Traffic
through the traditional human-driven probe vehicles, such 2.1. Simulation Settings
as global positioning system- (GPS-) based data and cell-
phone-based data. Traffic state estimation based on these 2.1.1. Assumptions. *e highly or fully automated CAVs
probe vehicles is one of the most effective methods because referring to Level 4 or Level 5 in the SAE autonomy level
probe vehicles have a wide coverage over space and time definitions [26] are still in development or test. *e simu-
[17–21]. *ere are two common categories of traffic state lation method provides a possibility to studying the mixed
estimation methods, i.e., the model-based methods and the traffic condition with CAVs. *ere is a bunch of expected
data-driven methods. *e model-based methods are made types of highly or fully automated CAVs. Different types of
up of two parts. *e first part is the traffic flow model, such as CAVs would lead to different influences on traffic. *erefore,
the Lighthill-Whitham-Richards (LWR) model [22], Payne this study made some preceding assumptions to clarify the
model [23], and their successors. *e second part is a data studied object and situation.
assimilation method to realize the estimation, such as First, CAVs are supposed to behave more assertively
Kalman filtering (KF) and its extensions [19, 24]. *e data- than RVs, and thus, they can maintain a shorter distance
driven methods mine the relationship between estimates and with the front vehicle.
observations from the historical big data. *e commonly Second, CAVs have a stronger ability to sense the traffic
used data mining techniques include the statistical analysis environment compared with RVs. *is ability could be en-
algorithms for the time-series data and the artificial intel- hanced either by the communication with everything
ligence models [20]. However, it should be noted that the (roadside unit, other vehicles, traffic management center, and
traditional probe-based methods are under the human- so on) or by the advance sensing facilities. As a result, this
driven mode, and the probe and non-probe vehicles are sensing range is supposed to be within ± 500m in this study.
supposed to have similar driving behaviour. As mentioned *ird, since this study is based on the simulation, the
before, the driving behaviours of CAVs are expected to be latency and packet loss of the communication between
different from those of RVs, so the applicability of the tradi- CAVs and everything (roadside unit, other vehicles, traffic
tional probe-based estimation methods is uncertain. *e data- management center, and so on) would not be considered this
driven methods require vast amount of historical data. time.
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 3

2.1.2. Simulation Parameters. *is study uses VISSIM 3. Discussions on Mixed Traffic
(version 9) to simulate the mixed traffic containing CAVs
and RVs. PTV Group has stated that CAV behaviour could *e 6 km mainline is divided by 500 m into 12 links. *ose
be modelled using VISSIM internally or externally [27]. *is links are then labelled from Link 1 to Link 12 same as the
study implements the internal way, which is to modify the travel direction, as shown in Figure 1. Data are integrated by
VISSIM default driving behaviour parameters. Compara- the time interval of one minute. *e average ground-truth
tively speaking, the internal way is simpler and more con- link speeds could be calculated by the ratio of the link length
venient to use, whereas the external approach is used when to the average travel time of all vehicles. *e average speed of
researchers want to define their own driving behaviour CAVs on a link during a time interval is calculated using the
models in VISSIM. Since the focus of this study is to estimate position and the timestamps of CAVs. Here are some sta-
speeds from data generated by CAVs with a low penetration tistical findings about the simulated mixed traffic.
rate, the internal way is more suitable and achievable.
PTV Group has given some recommendations to set the
internal model by changing the car-following and lane- 3.1. Sample Size and Data-Missing Rate. *e boxplot in
changing behaviour parameters for the CAVs [28]. In ap- Figure 2 shows the distribution of the sample size per minute
plication, there have been some works that are based on the under different CAV penetrations. *e median sample size
internal model in VISSIM to explore the impact of CAVs. per minute under the penetration rate of 1%, 3%, 5%, 7%,
Table 1 summarizes their adjusted parameters as well as the and 10% is 1, 2, 4, 5, and 7, respectively. When the pene-
corresponding default value in VISSIM 9. It should be noted tration rate is 1%, sample size per minute would mostly
that both this study and the works in Table 1 use the appear as a number within [1, 2]12. Similarly, it can be seen
Wiedemann 99 model as the car-following model for the that the most frequent sample size for the penetration rate of
freeway traffic. 3%, 5%, 7%, and 10% is [1, 3], [2, 5], [3, 7], and [5, 10],
Since no empirical data are available, these applications respectively. Besides, it seems that the variation of the
have indicated the possibility of modelling CAVs in VISSIM sample size adds with the increase of penetration rate.
internally, to some extent. Although it seems that each study Except for the sample size, another very concerned issue
has made different adjustment to the default values, they in the discussion of traffic probe with low penetration rate is
have something in common. For instance, they would let the the missing data rate. *is study defines the data-missing
CAV keep a shorter distance with the front vehicle, have rate on a link as the rate between the number of time in-
faster and smoother reactions, observe more around vehi- tervals that have collected CAV data and the total number of
cles, and realize the cooperative lane changing. Some dif- time intervals. Figure 3 presents the data-missing rate on
ferences might be caused by the different versions of different links and under different CAV penetrations.
VISSIM. For example, the maximum speed difference is Figure 3 shows that if the CAV penetration rate is small,
different between VISSIM version above 9 and below 9. the sample size is really small and there will be a serious data
Within the threshold present in these existing studies, this loss. Especially, when the proportion of CAVs is 1%, the
study made the following modifications to the internal data-missing rate almost reaches fifty percent. It requires
models in VISSIM 9, as shown in Table 2. RVs use the default that the estimation method is capable of filling the missing
values, while some parameters are adjusted for CAVs. Be- parts.
sides, the desired speed is reset as well, which is 80 km/h for
RVs and 90 km/h for CAVs. 3.2. Speed Difference. Afterwards, this study looks into the
difference between the average speeds of CAVs and the
2.2. Simulation Scenarios. A simplified freeway is simulated, mixed traffic speeds on a link. *is difference is calculated by
which contains a 6-km three-lane mainline in one travel the following equation:
direction, a one-lane on-ramp, and a one-lane off-ramp, as dij � vcij − vij , (1)
shown in Figure 1. *e simulation duration is 15300 s with a
900 s warm-up period. Data collected from 900 s to 15300 s where dij is the difference between the average speed of
are used for analysis. CAVs and the average link speed at the ith time interval and
To analyse the impact from CAV penetration rates, this jth link, vcij is the average speed of CAVs at the ith time
study proposes six scenarios with different compositions of interval and jth link, and vij is the average link speed at the
RVs and CAVs, as shown in Table 3. To indicate a traffic ith time interval and jth link.
condition with a low proportion of CAVs, the largest ratio of Table 5 summarizes the speed differences and their
CAVs in mixed traffic is set as 10%. In each scenario, the variances. *e desired speed of CAV is higher than that of
mixed traffic is loaded on mainline and on-ramp, which is RV, so it could be referred that the average speeds of CAVs
varying over time, as shown in Table 4. *e input traffic is set would be most likely higher than the average link speeds. It is
to approach the designed freeway lane capacity from the proved by the average and median speed differences in
simulation time 8100 s and last to 10700 s. Besides, in all Table 5. *e maximum and minimum differences indicate
scenarios, 15% of mainline traffic is assigned to leave the that the average speeds of CAV might also overestimate and
freeway at off-ramp. underestimate the link speeds. It would be vital to establish a
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 Journal of Advanced Transportation

Table 1: Existing studies on CAV simulation by the internal model.

Default Atkins Aria et al. Tibljaš et al. Stanek et al. Sukennik Asadi
Parameters
value [29] [30] [31] [32] [28] et al. [33]
VISSIM
Version of VISSIM VISSIM 9 VISSIM 8 DK DK VISSIM 9 VISSIM 11
10
CC0, standstill distance (m) 1.50 — — 1.00 1.25 1.00 1.00–1.50
CC1, headway time (gap between
0.90 — 0.30 0.50 0.25 0.60 0.50–1.50
vehicles) (s)
CC2, car-following distance/following
4 0 — 1.00 3 0 0
variation (m)
CC3, threshold for entering following
−8 — — — −12 −6 −8
(s)
CC4, negative following threshold
−0.35 0 −0.56 −0.10 — −0.10 0
(m/s)
CC5, positive following threshold
0.35 0 0.56 — — 0.10 0
(m/s)
CC6, speed dependency of oscillation
11.44 0 — 0 0 0 0
(1/(m/s))
CC7, oscillation during acceleration
0.25 — — 0.40 — 0.10 0.15–0.45
(m/s2)
2
CC8, standstill acceleration (m/s ) 3.50 — — 4 — 4 3.30–3.90
CC9, acceleration at 80 km/h (m/s2) 1.50 — — 2 — 2 1.30–1.90
Look-ahead distance 0 to 250 m — 150–200 m 0–105 m 0–500 m 0–250 m 0–800 m
Look-back distance 0 to 150 m — 150–200 m 0–280 m 0–250 m 0–150 m 0–800 m
Observed vehicles 2 10 6–8 — 10 99 10
Smooth close-up behaviour No — — Yes — — Yes
Accepted deceleration, trailing vehicle
−1 — — — — −1.50 —
(m/s2)
Minimum headway, front/rear (m) 0.50 — — — 0.40 0.50 0.20–0.70
Safety distance reduction factor 0.60 — — — 0.45 0.75 0.30–0.80
Maximum deceleration for cooperative
−3 — — — −4 −6 —
braking (m/s2)
Cooperative lane change No — — — Yes Yes Yes
Maximum speed difference (km/h) 3 — 3 — — 10.80 10.80

right relationship model between speeds of CAVs and link Besides, the critical speed to identify the free-flow state
speeds. Accordingly, the variance of speed differences is seems to remain the same at 80 km/h, as shown in Figure 4.
calculated as shown in Table 5, which could be applied to Since the lowly penetrated CAVs do not have a significant
calibrate the relationship model. *e variance shows the impact on the critical speed and volume, this study would
expectation of the squared deviation of the speed difference assume that the traditional estimation method (i.e., Kalman
from its mean difference, and Table 5 indicates the deviation filtering-based estimation method) might be effective when
decreases with the increase in the CAV penetration rate. the proportion of CAV in the mixed flow is low.

3.3. Fundamental Diagram of Mixed Flow. *is section aims 4. Kalman Filtering-Based Estimation
to discuss the impact of CAVs on the fundamental diagram. Method for Mixed Traffic
Taking Link 7 as an example, Figure 4 shows the speed-flow
4.1. Basic Kalman Filtering Algorithm. *e traditional Kal-
diagrams under different penetration rates of CAVs. It seems
man filtering-based estimation method is applied. For ap-
that an increase in the CAV penetration rate has a slight
plication in this study, the discrete form of the KF in the
impact on the shape of the speed-flow fundamental diagram.
linear speed model is given by
*e largest traffic volume (approaching the link capacity) is
7860, 7620, 7980, 7560, 7920, and 8040, respectively, when xt � At xt−1 + ωt , (2)
CAVs account for 0%, 1%, 3%, 5%, 7%, and 10%. It indicates
that an increase in the CAV penetration rate would not y t � H t xt + v t , (3)
definitely contribute to the increase in traffic flux, when the
penetration rate is under 10%. Moreover, when the pene- where xt is the average link speed at the tth time interval. For
tration rate increases, the number of the scatter dots on the simplicity, it is originally assumed that it has a linear re-
left side reduces. To some extent, it indicates that the in- lationship with the speed value at the previous time interval.
crease of CAVs in the mixed flow could relief the traffic yt is the collected speed which is the average speed of CAVs
congestion. at the tth time interval. Similarly, the CAV speed is supposed
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 5

Table 2: Car-following and lane-change parameters of RVs and CAVs in this study.
Parameters Default value for RVs Adjusted value for CAVs
CC0, standstill distance (m) 1.5 1.25
CC1, headway time (gap between vehicles) (s) 0.9 0.5
CC2, car-following distance/following variation (m) 4 3
CC3, threshold for entering following (s) −8 −12
CC4, negative following threshold (m/s) −0.35 −0.1
CC5, positive following threshold (m/s) 0.35 0.1
CC6, speed dependency of oscillation (1/(m/s)) 11.44 0
CC7, oscillation during acceleration (m/s2) 0.25 0.25
CC8, standstill acceleration (m/s2) 3.5 3.5
CC9, acceleration at 50 miles per hour (m/s2) 1.5 1.5
Look-ahead distance 0 to 250 m 0–500 m
Look-back distance 0 to 150 m 0–500 m
Observed vehicles 2 10
Smooth close-up behaviour Checked Checked
General behaviour Free lane selection Free lane selection
Maximum deceleration, own vehicle (m/s2) −4 −4
Maximum deceleration, trailing vehicle (m/s2) −3 −3
−1 m/s2 per distance, own vehicle and trailing vehicle (m) 100 100
Accepted deceleration, own vehicle (m/s2) −1 −1
Accepted deceleration, trailing vehicle (m/s2) −1 −1
Waiting time before diffusion (s) 60.00 60.00
Minimum headway, front/rear (m) 0.5 0.37
Safety distance reduction factor 0.6 0.45
Maximum deceleration for cooperative braking (m/s2) −3 −4
Cooperative lane change Not checked Checked
Maximum speed difference (km/h) 3 3
Maximum collision time (s) 10 10
Collision time gain (s) 2 2
Minimum longitudinal speed (km/h) 1 1
Time before direction changes (s) 0 0
Overtake same lane vehicle, minimum lateral distance standing (m) at 0 km/h 1 0.75
Overtake same lane vehicle, minimum lateral distance driving (m) at 50 km/h 1 0.75

1500m 3000m 1500 m

Travel direction Off
p -ra
ram mp
On-

Figure 1: Simulated freeway network diagram.

Table 3: Simulation scenarios of traffic composition.

Scenarios Proportion of RVs (%) Proportion of CAVs (%)
1 100 0
2 99 1
3 97 3
4 95 5
5 93 7
6 90 10

to have a linear relationship with the average link speed. At equation (3) describes how the state vector is related to an
and Ht are the linear coefficients. ωt and vt represent the m-dimensional measurement vector yt . Obviously, in this
state and measurement noises, respectively. Usually, study, m and n are mostly not equal. Especially, when
E(ωt ) � 0, E(vt ) � 0, E(ωs , ω′t) � Qt , E(vs , v′t) � Rt , and CAVs account for 1%, m is far less than n. In the presence
E(ωs , v′t) � 0. *e state equation (2) shows the behaviour of of incomplete data, the following recursive formula is
an n-dimensional state vector xt , and the measurement used to solve the previous discrete model in this study:
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 Journal of Advanced Transportation

Table 4: Traffic input in a scenario.

Simulation time interval (s) Mainline traffic input (three-lane) (veh) On-ramp traffic input (one-lane) (veh)
0∼2700 3000 300
2700∼4500 3600 300
4500∼6300 4500 300
6300∼8100 5100 300
8100∼9900 6000 300
9900∼11700 6300 300
11700∼12600 5100 300
12600∼13500 4500 300
13500∼14400 3600 300
14400∼15300 3000 300

e distribution of CAV sample size

22
21 +
20 +
19 +
18 + +
17 +
Number of CAVs per minute

16 +
15 +
14 +
13 +
12 +
11 +
10 + +
9 +
8 + +
7 + +
6 +
5 +
4 +
3
2
1
0
1 3 5 7 10
Penetration rate of CAVs (%)

Number of CAVs under 1% penetration

Number of CAVs under 3% penetration
Number of CAVs under 5% penetration
Number of CAVs under 7% penetration
Number of CAVs under 10% penetration
Figure 2: Distribution of sample size under different CAV penetration rates.

⎪
⎧
⎪ Pt−1 ′
t · Ht
⎨ 􏽢
x t−1
+ 􏽢 t−1
􏼐y − H t · x t 􏼑, if yt is collected,
t
􏽢t � ⎪
x
t
Ht · Pt · Ht + Rt t
t−1 ′
⎪
⎩ t−1
􏽢t ,
x if yt is missing,
(4)
⎧
⎪
⎪ Pt−1 ′
t · Ht · Ht
t ⎨ Pt−1
t 􏼠1 − t−1 · H′ + R 􏼡,
if yt is collected,
Pt � ⎪ H t · P t t t
⎪
⎩ t−1
Pt , if yt is missing,

where Ptt is the error covariance matrix of state xt and example shown in Figure 5, it is obvious that the vari-
Pt−1
t � At · Pt−1 ′
t−1 · At + Qt . Usually, At , Ht , Qt , and Rt are ation of state error enlarges when speed falls below the
calibrated using the historical data. According to the critical speed of 80 km/h. From the observations among
small average speed difference in Table 5, this study sets all fundamental diagrams across all penetration rates, the
both At and Ht as 1. As for the state noise Qt , this study application of CAVs does not have a significant impact
calibrates it separately based on the traffic condition. on the critical speeds when CAVs have a low penetration
With a mixed traffic of 1% CAV penetration rate as an rate. *erefore, the same critical speed is used to identify
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 7

0.6

0.5

0.4

Missing rate 0.3

0.2

0.1

0
1 2 3 4 5 6 7 8 9 10 11 12
Link no.

1% 7%
3% 10%
5%
Figure 3: Data-missing rate on different links.

Table 5: Speed difference.

Link no.
Speed difference CAV penetration rate (%)
1 2 3 4 5 6 7 8 9 10 11 12
1 3 1 1 2 2 1 1 1 2 3 2 2
3 3 2 2 2 2 2 1 1 1 3 3 3
Average difference (km/h) 5 3 2 2 2 2 2 2 2 2 4 3 3
7 3 2 2 2 2 2 2 2 2 4 3 3
10 3 2 3 2 2 2 2 2 2 3 3 3
1 3 1 1 1 2 1 1 1 2 2 1 1
3 3 1 1 1 1 1 1 1 1 3 2 2
Median difference (km/h) 5 3 2 2 2 1 1 1 1 1 4 3 3
7 3 2 2 2 2 1 1 1 1 3 3 3
10 3 2 2 2 2 2 2 1 2 3 3 3
1 21 32 12 17 29 19 19 20 22 18 14 14
3 16 16 15 22 27 19 16 27 23 23 27 18
Maximum difference (km/h) 5 14 11 16 16 13 22 20 22 22 21 17 14
7 14 12 15 16 17 21 19 19 16 18 16 15
10 15 17 14 13 26 15 16 18 22 15 23 14
1 −10 −24 −13 −14 −8 −9 −20 −13 −12 −8 −7 −6
3 −9 −6 −8 −12 −9 −9 −10 −15 −39 −9 −4 −6
Minimum difference (km/h) 5 −3 −5 −8 −8 −4 −9 −12 −15 −12 −12 −9 −5
7 −6 −5 −4 −8 −4 −5 −8 −21 −20 −4 −6 −3
10 −3 −5 −5 −6 −6 −14 −10 −5 −18 −7 −4 −4
1 23 32 20 30 27 20 26 27 33 22 19 17
3 14 13 14 27 18 16 19 24 41 18 22 19
Variance (km/h)2 5 9 8 11 14 11 13 14 18 20 16 14 13
7 8 7 7 12 9 10 12 16 19 12 10 10
10 6 6 6 8 8 8 8 8 17 8 9 10

the traffic state. *e state noise Qt is calibrated under 4.2. Estimation Results. Using the proposed KF-based esti-
free-flow condition and non-free-flow condition sepa- mation method, the speed estimates are obtained. Taking the
rately. *e measurement noise Rt could be calculated scenarios of 1% and 10% as an example, Figure 6 exhibits the
using the variance of difference between CAV speeds and ground-truth link average speeds, the average speeds of CAVs,
ground-truth speeds, as shown in Table 5. Finally, the and the estimated speeds on Link 7. First, the results indicate
initial state values are set as x0 = 85 and P0 = Qt under the that the estimation method interpolates the missing parts of
free-flow condition. CAV data. Second, the estimation method smooths and
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 Journal of Advanced Transportation

9000 9000
8000 8000
Traffic flow (veh/h)

Traffic flow (veh/h)

7000 7000
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Average link speeds (km/h) Average link speeds (km/h)
(a) (b)
9000 9000
8000 8000
Traffic flow (veh/h)

Traffic flow (veh/h)

7000 7000
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Average link speeds (km/h) Average link speeds (km/h)
(c) (d)
9000 9000
8000 8000
Traffic flow (veh/h)

Traffic flow (veh/h)

7000 7000
6000 6000
5000 5000
4000 4000
3000 3000
2000 2000
1000 1000
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Average link speeds (km/h) Average link speeds (km/h)
(e) (f )

Figure 4: Speed-flow diagrams on Link 7. (a) No CAVs. (b) CAVs account for 1%. (c) CAVs account for 3%. (d) CAVs account for 5%.
(e) CAVs account for 7%. (f ) CAVs account for 10%.

CAVs account for 1%

100.00 40.00

30.00
80.00
20.00
Speed difference
Speed (km/h)

60.00 10.00

40.00 0.00

–10.00
20.00
–20.00

0.00 –30.00
1
11
21
31
41
51
61
71
81
91
101
111
121
131
141
151
161
171
181
191
201
211
221
231

Time interval no.

Link speed at ith time interval

Speed difference between (i – 1)th and ith time interval
Figure 5: An example for speed difference between (i–1)th and ith time interval.
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 9

120

100

80

Speed (km/h) 60

40

20

0
1
10
19
28
37
46
55
64
73
82
91
100
109
118
127
136
145
154
163
172
181
190
199
208
217
226
235
Time interval no.

Ground truth
CAV
Estimate
(a)
120

100

80
Speed (km/h)

60

40

20

0
1
10
19
28
37
46
55
64
73
82
91
100
109
118
127
136
145
154
163
172
181
190
199
208
217
226
235 Time interval no.

Ground truth
CAV
Estimate
(b)

Figure 6: Ground-truth speeds, CAV speeds, and estimated speeds on Link 7. (a) CAVs account for 1%. (b) CAVs account for 10%.

modifies the CAV speeds, and thus, the estimated speeds are 4.3. Accuracy. *is section will further evaluate the esti-
closer to the ground truth. mation accuracy. It is measured by RMSE (Root Mean
Furtherly, the ground-truth speeds, estimated speeds, Squared Error) and MAE (Mean Absolute Error). *ey can
and CAV speeds are illustrated in the time-space form, as be obtained by
shown in Figure 7. According to the speed values, traffic state 􏽶�������������
􏽴
is divided into three conditions which are represented by 1 m
three different colors, i.e., green, yellow, and red. Figure 7 RMSE � 􏽘 x −x 􏽢 i 􏼁2 ,
m i�1 i
shows that the estimates (i.e., Figures 7(a) and 7(d)) almost (5)
copy the ground-truths (i.e., Figures 7(b) and 7(e)). If not for
the data missing, CAV speeds could almost tell the traffic 1 m 􏼌􏼌􏼌 􏼌􏼌
MAE � 􏽢 i 􏼌􏼌.
􏽘􏼌x − x
condition, as shown in Figures 7(c) and 7(f ). Especially, 10% m i�1 i
CAVs (i.e., Figure 7(f )) seem to be able to visualize the traffic
state in a rough three-color map compared with the ground- Figure 8 presents the RMSE and MAE of estimated
truth speed map (i.e., Figure 7(e)). speeds at each penetration rate. Comparatively, the scenario
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 Journal of Advanced Transportation

240 240 240

90
220 220 220 100
80
200 80 200 200
180 70 180 70 180 80

Time interval no.

Time interval no.
Time interval no.

160 160 160

60 60
140 140 140
50 60
120 120 120
50
100 40 100 100
80 40
80 30 40 80
60 60 60
20 30 20
40 40 40
20 10 20 20 20
Absent
2 4 6 8 10 12 2 4 6 8 10 12 2 4 6 8 10 12
Link no. Link no. Link no.

(a) (b) (c)

240 240 240
220 90 220 220 100
80
200 200 200
180 80 180 180
70 80

Time interval no.

Time interval no.
Time interval no.

160 70 160 160

140 140 60 140
60 60
120 120 120
100 50 100 50 100
80 40
80 80
60 40 60 40 60
40 40 40 20
30 30
20 20 20
20 Absent
2 4 6 8 10 12 2 4 6 8 10 12 2 4 6 8 10 12
Link no. Link no. Link no.

(d) (e) (f )

Figure 7: Estimated speeds, the ground-truth speeds, and the CAV speeds in the color map. (a) Estimated speeds with 1% penetration rate of
CAVs (km/h). (b) Ground-truth speed when CAVs account for 1% (km/h). (c) CAV speeds with 1% CAVs in mixed traffic (km/h). (d)
Estimated speeds with 10% penetration rate of CAVs (km/h). (e) Ground-truth speeds when CAVs account for 10% (km/h). (f ) CAV speeds
with 10% CAVs in mixed traffic (km/h).

7 5
6.5 4.5
6
5.5 4
5 3.5
RMSE

MAE

4.5
4 3
3.5 2.5
3
2.5 2
2 1.5
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Link no. Link no.

1% 7% 1% 7%
3% 10% 3% 10%
5% 5%
(a) (b)

Figure 8: (a) RMSE and (b) MAE of estimated speeds.

with 10% CAVs has smaller RMSE and MAE than other data that eliminates the data-missing time intervals. Taking
penetration scenarios. However, RMSE and MAE of the 3%, the data from Link 7 as an example, the accuracy comparison
5%, and 7% scenarios are quite approaching those of the 10% results are shown in Figure 9. It is obvious that estimates
scenario. Although RMSE and MAE of the 1% scenario are a reduce the speed error compared with the CAV speeds.
bit larger than other scenarios, their values remain within a
small value, i.e., RMSE is less than 7 and MAE is less than 5. 4.4. Sensitivity Analysis. In the application of this KF-based
In general, the estimation method with limited CAV data has estimation method, some parameters might play an im-
a reasonable performance, even when the proportion of portant role in the estimation accuracy. *ey are the state
CAVs in mixed traffic is only 1%. and measurement noises. As mentioned in the KF-based
Moreover, this study would like to compare the accuracy estimation method, the measurement noise is calibrated by
of estimates and CAV speeds. Since there are missing parts the historical data of the CAV speeds and the ground-truth
in CAV speeds, RMSE and MAE are calculated using the speeds on each link. In practice, the ground-truth speeds are
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 11

0
1% 3% 5% 7% 10%
CAV penetration rate

RMSE-CAV MAE-CAV
RMSE-estimate MAE-estimate
Figure 9: Accuracy comparison between estimated speeds and CAV speeds.

5.50 3.80
3.60
5.00
3.40
4.50
3.20
RMSE

MAE

4.00 3.00
2.80
3.50
2.60
3.00
2.40
2.50 2.20
1% 3% 5% 7% 10% 1% 3% 5% 7% 10%
CAV percentage CAV percentage

R-calibrated R-calibrated
R-min R-min
R-max R-max
(a) (b)

Figure 10: *e impact of measurement error (R) on accuracy. (a) Impact on RMSE. (b) Impact on MAE.

RMSE MAE

5.39 3.76
4.88 3.30
3.79 3.71 2.87 2.73
3.26 3.35 3.36 2.51 2.45 2.51 2.54
3.20 2.97 2.26 2.33
2.83

1% 3% 5% 7% 10% 1% 3% 5% 7% 10%
CAV percentage CAV percentage

Q changing with traffic state Q changing with traffic state

Q unchanged Q unchanged
(a) (b)

Figure 11: *e impact of state error (Q) on accuracy. (a) Impact on RMSE. (b) Impact on MAE.
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 Journal of Advanced Transportation

usually not available on all links. *erefore, this study selects Conflicts of Interest
out the minimum and maximum Rt from Table 5 and uses
these values in the estimation. *eir estimation accuracy is *e authors declare that there are no conflicts of interest
compared with those using the calibrated Rt , as shown in regarding the publication of this paper.
Figure 10. Obviously, the proposed method is the optimal,
but if Rt could not be calibrated on each link, a small value of Acknowledgments
Rt calibrated on other links is suggested.
Another parameter is the state noise Qt . In the proposed *e authors would also like to thank the simulation software
method, both the calibration and application of the state noise support from Texas A&M University. *is research was
would be separated based on the traffic condition. If this supported by the Natural Science Foundation of Jiangsu
separation is eliminated, this study finds that it will lead to Province of China (Grant no. BK20180486), China Post-
larger estimation errors as the comparison results in Figure 11. doctoral Science Foundation (Grant no. 2018M642257), the
It indicates the outperformance of the proposed method. Fundamental Research Funds for the Central Universities
(Grant no. 330920021140), Key Laboratory for Automotive
Transportation Safety Enhancement Technology of the
5. Conclusions and Future Works Ministry of Communication (Grant no. 300102229506),
It seems to be inevitable that CAVs will come into the National Social Science Fund of China (Grant no.
market and travel on the regular roads in the near future. It 18CFX062), and National Key R&D Program Intergov-
also could be imaged that there will be mixed traffic ernmental International Science and Technology Innovation
consisting of CAVs and RVs, and the proportion of CAVs Cooperation Key Project (Grant no. 2016YFE01018000).
will be low at the beginning stage. *is study discussed the
application of the limited CAV data to estimate traffic state References
at this beginning stage. At first, this study set up a [1] R. Hoogendoorn, B. V. Arem, and S. Hoogendoorn, “Auto-
microsimulation platform of the mixed traffic flow using mated driving, traffic flow efficiency, and human factors:
the VISSIM. Five simulation scenarios with the CAV literature review,” Journal of the Transportation Research
penetration increasing from 1% to 10% were set to generate Board, vol. 2422, no. 1, pp. 113–120, 2014.
the testing data. *en, a step-by-step discussion on the [2] S. Childress, B. Nichols, B. Charlton, and S. Coe, “Using an
characteristics of mixed traffic was conducted based on the activity-based model to explore the potential impacts of au-
simulation data. *e sample size distribution under dif- tomated vehicles,” Transportation Research Record: Journal of
ferent CAV penetrations was found, and the data-missing the Transportation Research Board, vol. 2493, no. 1, pp. 99–
rate was calculated which was especially large when CAVs 106, 2015.
only account for 1% of mixed traffic. *e analysis on the [3] D. Milakis, B. V. Arem, and B. V. Wee, “Policy and society
related implications of automated driving: a review of liter-
speed difference between CAV speeds and the ground-
ature and directions for future research,” Journal of Intelligent
truth link speeds was an assistant in the following cali- Transportation Systems: Technology, Planning, and Opera-
bration of the proposed estimation model . *e speed-flow tions, vol. 21, no. 4, pp. 324–348, 2015.
diagrams of mixed traffic indicated the possibility of ap- [4] P. Bansal and K. M. Kockelman, “Forecasting Americans’
plying the traditional estimation method. Accordingly, the long-term adoption of connected and autonomous vehicle
simple KF-based estimation method was used and adjusted technologies,” Transportation Research Part A: Policy and
to adapt the incomplete CAV data. *e estimation results, Practice, vol. 95, pp. 49–63, 2017.
accuracy evaluations, and sensitivity analysis validated the [5] T. Litman, “Autonomous vehicle implementation predictions:
applicability and precision of the proposed estimation implications for transport planning,” in Proceedings of the
method using limited CAV data. Transportation Research Board 94th Annual Meeting,
Washington, DC, USA, January 2015.
Since the Level 4 and Level 5 CAVs are not ready to enter
[6] L. C. Davis, “Effect of adaptive cruise control systems on
the market, the simulation method is an alternative way to traffic flow,” Physical Review E, vol. 69, no. 6, Article ID
make these investigations. With the developing technology 066110, 2004.
of CAVs, the driving behaviour model of CAVs needs to be [7] S. Shladover, D. Su, and X.-Y. Lu, “Impacts of cooperative
updated accordingly in the future. Besides, the simulated adaptive cruise control on freeway traffic flow,” Journal of the
roadway is oversimplified. *e complex merging and Transportation Research Board, vol. 2324, pp. 63–70, 2012.
weaving area from on/off-ramp to the mainline could be [8] B. Friedrich, “*e effect of autonomous vehicles on traffic,” in
discussed in detail. Moreover, the fusion of the measure- Autonomous Driving, Springer, Berlin, Germany, 2016.
ments from other existing sensors, such as loop detectors [9] M. Zhou, X. Qu, and S. Jin, “On the impact of cooperative
and GPS probe vehicles, and the testing of other existing autonomous vehicles in improving freeway merging: a
estimation methods are also useful for the field application. modified intelligent driver model-based approach,” IEEE
Transactions on Intelligent Transportation Systems, vol. 18,
no. 6, pp. 1422–1428, 2017.
Data Availability [10] L. Xiao, M. Wang, W. Schakel, and B. van Arem, “Unravelling
effects of cooperative adaptive cruise control deactivation on
*e data used to support the findings of this study were traffic flow characteristics at merging bottlenecks,” Trans-
generated from VISSIM simulation software based on the portation Research Part C: Emerging Technologies, vol. 96,
simulation settings described within the article. pp. 380–397, 2018.
 1409, 2020, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1155/2020/1361583 by HEALTH RESEARCH BOARD, Wiley Online Library on [14/03/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Advanced Transportation 13

[11] L. D. Baskar, B. D. Schutter, and J. Hellendoorn, “Optimal https://web.archive.org/web/20170903105244/https://www.

routing for intelligent vehicle highway systems using a sae.org/misc/pdfs/automated_driving.pdf.
macroscopic traffic flow model,” in Proceedings of the 12th [27] B. Cisco, “PTV VISSIM & connected autonomous vehicles,”
International IEEE Conference on Intelligent Transportation in Proceedings of the 18th Annual North America PTV Vision
Systems, pp. 576–581, St. Louis, MO, USA, October 2009. Traffic User Group Meeting, Portland, OR, USA, 2017.
[12] Y. Liu, J. Guo, J. Taplin, and Y. Wang, “Characteristic analysis [28] P. Sukennik, Micro-Simulation Guide for Automated Vehicles,
of mixed traffic flow of regular and autonomous vehicles using 2018, https://www.h2020-coexist.eu/wp-content/uploads/
cellular automata,” Journal of Advanced Transportation, 2018/11/D2.5-Micro-simulation-guide-for-automated-vehicles.
vol. 2017, Article ID 8142074, 10 pages, 2017. pdf.
[13] C. Lu and A. Aakre, “A new adaptive cruise control strategy [29] Atkins, Research on the Impacts of Connected and Autono-
and its stabilization effect on traffic flow,” European Transport mous Vehicles (CAVs) on Traffic Flow, Atkins, Epsom, UK,
Research Review, vol. 10, no. 49, pp. 1–11, 2018. 2016.
[14] M. M. Morando, Q. Tian, L. T. Truong, and H. L. Vu, [30] E. Aria, J. Olstam, and C. Schwietering, “Investigation of
“Studying the safety impact of autonomous vehicles using automated vehicle effects on driver’s behavior and traffic
simulation-based surrogate safety measures,” Journal of Ad- performance,” Transportation Research Procedia, vol. 15,
vanced Transportation, vol. 2018, Article ID 6135183, 11 pages, pp. 761–770, 2016.
2018. [31] A. D. Tibljaš, T. Giuffrè, S. Surdonja, and S. Trubia, “Intro-
[15] C. Lu, J. Dong, L. Hu, and C. Liu, “An ecological adaptive duction of autonomous vehicles: roundabouts design and
cruise control for mixed traffic and its stabilization effect,” safety performance evaluation,” Sustainability, vol. 10, no. 4,
IEEE Access, vol. 7, pp. 81246–81256, 2019. p. 1060, 2018.
[16] C. Lu, J. Dong, and L. Hu, “Energy-efficient adaptive cruise [32] D. Stanek, E. Huang, R. T. Milam, and Y. Wang, “Measuring
control for electric connected and autonomous vehicles,” autonomous vehicle impacts on congested networks using
IEEE Intelligent Transportation Systems Magazine, vol. 11, simulation,” in Proceedings of the Transportation Research
no. 3, pp. 42–55, 2019. Board 97th Annual Meeting, Washington, DC, USA, 2018.
[17] J. C. Herrera and A. M. Bayen, “Incorporation of Lagrangian [33] F. E. Asadi, A. K. Anwar, and J. C. Miles, “Investigating the
measurements in freeway traffic state estimation,” Trans- potential transportation impacts of connected and autono-
portation Research Part B: Methodological, vol. 44, no. 4, mous vehicles,” in Proceedings of the 2019 IEEE International
pp. 460–481, 2010. Conference on Connected Vehicles and Expo (ICCVE), Graz,
[18] J. C. Herrera, D. B. Work, R. Herring, X. Ban, Q. Jacobson, Austria, November 2019.
and A. M. Bayen, “Evaluation of traffic data obtained via GPS-
enabled mobile phones: the mobile century field experiment,”
Transportation Research Part C: Emerging Technologies,
vol. 18, no. 4, pp. 568–583, 2010.
[19] S. He, Y. Cheng, F. Ding, J. Zhang, and B. Ran, “Extended
Kalman filter-based freeway traffic state estimation using
cellphone activity data,” in Proceedings of the Transportation
Research Board 95th Annual Meeting, Washington DC, USA,
January 2016.
[20] S. He, J. Zhang, Y. Cheng, X. Wan, and B. Ran, “Freeway
multisensor data fusion approach integrating data from
cellphone probes and fixed sensors,” Journal of Sensors,
vol. 2016, Article ID 7269382, 13 pages, 2016.
[21] S. He, Y. Cheng, G. Zhong, and B. Ran, “A data-driven study
on the sample size characteristics of cellular handoff probe
system,” Advances in Mechanical Engineering, vol. 9, no. 4,
pp. 1–9, 2017.
[22] M. J. Lighthill and G. B. Whitham, “On kinematic waves II. A
theory of traffic flow on long crowded roads,” Proceedings of
@e Royal Society A Mathematical, Physical and Engineering
Sciences, vol. 229, no. 1178, pp. 317–345, 1955.
[23] H. J. Payne, Models of Freeway Traffic and Control (Mathe-
matical Models of Public Systems), Simulation Councils, Inc.,
Vista, CA, USA, 1971.
[24] T. Seo, T. Kusakabe, and Y. Asakura, “Traffic state estimation
with the advanced probe vehicles using data assimilation,” in
Proceedings of the IEEE 18th International Conference on
Intelligent Transportation Systems, IEEE Computer Society,
Washington, DC, USA, 2015.
[25] R. Wang, Y. Li, and D. B. Work, “Comparing traffic state
estimators for mixed human and automated traffic flows,”
Transportation Research Part C: Emerging Technologies,
vol. 78, pp. 95–110, 2017.
[26] SAE, “Automated driving: levels of driving automation are
defined in new SAE international standard J3016,” 2017,