MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de

Eletromagnetismo

Hybrid Multicriteria Model for Return Channel

Selection in IDTV Environment
Marcos César da Rocha Seruffo, Ádamo Lima de Nandamudi Lankalapalli Vijaykumar
Santana, Carlos Renato Lisboa Francês National Institute for Space Research (INPE)
Laboratory of Network Planning for High Performance São José dos Campos – São Paulo
Federal University of Para vijay@[Link]
Belém – Pará
{seruffo,adamo,rfrances}@[Link]

Abstract – The purpose of this paper is present a hybrid production chain involving programs and applications
multicriteria model which combine multivariable linear regression, development of interactive character. Based on this aspect,
to define weights for criteria to decision making, with multicriteria service providers should be concerned with the diversity of
decision making model – TOPSIS and ELECTRE III. This applications and services that can be offered.
combination is to be used to select return channel from iDTV
Environment. This proposal extracts weight vector from knowledge One of the main challenges, for social and digital inclusion
the collected data from real scenarios, without the need of through Digital TV, is to enable the population, mainly located
specialists, as usually is the case with traditional methods. This in areas of difficult access and low-income, to gain access to
resulted in a generic and flexible model and has been validated with services such as t-health, t-education, as well as to interactivity
a case study that ordered the preferred technologies that have been (e.g. online shopping, chats, video conferencing services). To
evaluated: ADSL2+, 3G, WiMAX and PLC. this end, allowing such facilities to be available in homes with
Keywords— Intelligent Systems, Interactive Digital TV,
access to traditional broadcast TV characterizes a major
Multivariable Regression, Model Decision Making, Service challenge to access providers.
Provider and Return Channel. This challenge is closely related to the capacity/availability
of access networks, whose capillarity must be much larger with
I. INTRODUCTION
respect to what is now available for Internet services.
Interactive Digital TV environment allows a new class of
services around broadcast of television content, in this aspect, Thus, it is fundamental to address the issue of first mile,
service providers have to plan the implantation of technology, i.e., the first technology with which a user connects to a
based on new demands that will be part of computing telecommunications network; this term is used synonymously
environments. The diversity of broadcast applications and with technology access that ensures the interactivity (return
usage of an Interactive Channel, as external communication in channel) service. Studies must be conducted so that service
the Brazilian context, leads to a case-study with the objective to providers have an idea of system performance characteristics.
consider the social inclusion specified in Brazilian Presidential In order to achieve this, scenarios must be studied and
Decree No. 4901. understood so that applications embedded in SBTVD are
available to the end user, so that providers can anticipate
It’s even more necessary as the Brazilian Digital TV problems in the system.
System (SBTVD) presents as a promising alternative when
considering digital inclusion, since the television is present in Studies must consider combining features of access
more than 90% of Brazilian homes, according to National networks (return channel), quality of service (QoS)
Survey by Household Sample of Brazilian Institute of requirements associated with applications, degree of
Geography and Statistics (IBGE) [1]. It is also available to interactivity such applications impose on the system and
virtually all social classes, in all regions of Brazil, which is not possible strategies for planning the capacity of interactive
true in the case of other communication devices, such as services.
computers. As a contribution, this article proposes a provision of
Therefore, one can consider that Digital TV has a great service strategies for selecting return channel within interactive
scope and it should not be restricted to a mere process of digital TV (iDTV) environment. This approach is based on a
improving the quality of transmission. It must be considered as combination of multivariable regression with TOPSIS
a tool for digital inclusion, naturally with strong possibility of (Technique for Order Preference by Similarity to Ideal
interactivity. Solution) and ELECTRE III (Elimination and Choice
Translating Reality), enabling decision making from a set of
In this context, a universe of activities involving production QoS parameters. With respect to measures, test results are
of audio/visual content to include applications that allow the obtained for measuring the delay, throughput, jitter, response
user to interact with the interactive TV applications has been time, data dropped and number of retransmissions from access
growing. The possibility to offer applications to viewers by technologies.
means of Interactive Digital Television originates a new

153
 MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de
Eletromagnetismo

It is important to note that the proposed approach is generic which, in turn, will generate an increase in accessing
and flexible and can easily incorporate other parameters and telecommunication infrastructures, and thus leading to
technologies, providing new decision making for different congestion problems. In [8] and [9], the authors propose a
scenarios. method for adaptive congestion control for the intensive
telecommunication traffic expected in case of integrated
The article is organized as follows: the second section
communication to broadcasting services. A method was
shows published literature and the real contributions of this proposed, based on sample monitoring, to allow access
article; the third presents models for decision making, showing
control, for volatile traffic in interactive television services.
the methods adopted in this article; in the fourth section tests,
scenarios, characterization of load and simulation are One way to measure the load of the systems being used in
addressed; the fifth section shows the application of order to prevent bottlenecks is by using monitoring and
multivariable linear regression, and the extraction of the control tools. For this, it is important to comment on the
weights of the application; the sixth section contains tests to concept of multicriteria decision making methods.
evaluate the first miles; the seventh section presents the
The use of heuristics for multicriteria decision making has
heuristic used for multicriteria decision making and analyzes
the results of experiments, making the comparison between been widely used, as shown in [10], with several papers
addressing algorithms for selecting the return channel in
TOPSIS and ELECTRE III; the last section has the conclusions
of the authors. heterogeneous environments, such as [11], [12], [13], [14],
[15] e [16], with different kinds of models.
II. RELATED WORK In order to create a heuristic for the proposed decision
With a constantly changing environment, several studies, making, a technique is investigated to extract knowledge from
published in the areas of service provider, Digital TV, access the collected data from real scenarios. Linear regression
technologies and methods for decision making, show that estimates conditional value (expected value) of a variable y,
strategies for providing service to select the return channel is given values to other variables x. Multivariable Linear
an emerging field of research. Regression extends simple linear regression onside ring
multiple input or explanatory variables. According to [17]
Digital TV systems can be represented using a layered MLR may be established through explanatory variables X,
architecture where each layer provides services for the upper with m columns and n lines. With a column vector consisting
layer and uses services from the lower layer. The of resulting variable y, 1 column and n lines, it is expected to
representation of the architecture is the same for all systems establish a linear relationship X and y. So, regression deals
and the differences are shown in different parts of the with estimating a expected conditional value.
architecture related to the process of modulation, coding,
compression, transmission, running applications and adopted Decision making is used to select a best option from pre-
middleware. The latter, in the Brazilian case, is called Ginga. established criteria. Analytical methods consider criteria, with
support from decision maker and aids in decision making
Several studies address the issue of how to connect the process. Criteria characterize the system and each criterion has
end-user to the telecommunications networks, as reviewed by an associated weight that informs the importance of that
[2]. Authors in [3] address the concern in the behavior of criterion. All the associated weights form an array of weights.
networks given the increase in the number of end users with
access to systems like ADSL (Asymmetric Digital Subscriber A domain specialist, usually, assigns the array of weights,
Line), cable modem, wireless, PLC (Power Line by analyzing the established criteria and evaluating the degree
Communication) and optical fiber. of their importance. The academic community accepts this sort
of definition. However, different results are obtained based on
In [4] The results provided evidence to determine the different specialists’ opinions and results differ very much
suitability of WiMAX (Worldwide Interoperability for even the criteria are the same [18].
Microwave Access) as a return channel for most interactive
applications of digital TV; In [5] are used a generic So, the objective of the work presented in this paper is to
architecture that combines PLC as a return channel for employ MLR to define array of weights in decision making.
interactive digital television. TOPSIS and ELECTRE III analytical methods of multicriteria
decision making were used. The combined use of linear
There is also research that considers using 3G access regression with multicriteria methods for decision making is
technology for digital TV. In [6] is investigated end-to-end quite new. So, very few publications report this topic and
performance adopting 3G as return channel. The motivation those avaliable leave several gaps1 that must be observed.
was to understand the behavior of radio technology as return
channel for an interactive Digital TV. In [19], the authors propose a regression approach for
estimating the decision weights of AHP using linear mixed
Among the range of access technologies available in the models (LMM). Other than determining the weight vectors,
market, the selection of access technology to be used by the this model also allows us to decompose the variation or
service provider is the subject much discussed in the literature. uncertainty in experts' judgment. In particular, the variation
As in [7], that uses the stochastic control technique, Markov among experts and the residual uncertainty due to rounding
Decision Process (MDP), studies and examines the errors in AHP scale or due to inconsistency within individual
relationship between optimal decisions that should be applied expert's judgments can be estimated and rigorously tested
by the service provider. using well-known statistical theories.
Another aspect found in the literature is related to the
1
increase in the number of interactive digital television devices, The term gap refers to a certain research has ceased to be considered in a
paper.

154
 MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de
Eletromagnetismo

The combined use of linear regression with multicriteria conversation via a spy program network, and Table I
methods for decision making is new, and the existing studies summarizes the measured statistical results of this application.
don’t include some gaps that need attention, to ensure the TABLE I. VALUES DTV-EDUC 2.0
proposed heuristic, which include: the lack of studies using Parameter Value
heuristics, based on multicriteria analysis, using data collected Packets Transmitted 575
in real scenarios; testing implementations on real devices; Average Packet per Second 1.016
implementation of heuristics in open platforms; using Average Size of Package 59.927 bytes
Multivariate Linear Regression to define the weight vector; Average Throughput 60.898 bytes/s
using the TOPSIS and ELECTRE III methods from a weight The measured data in Table I were used for configuring the
vector based on simulation tests, not experts opinion; application in a simulation scenario. It is noteworthy that this
comparison of different technologies that are already application when compared with other types of traditional
consolidated in the market. applications doesn’t occupy the network very much, because
III. TESTS its traffic is only in plain text that is typed by users. In general,
applications used in digital TV environment present this
Once the models that would be used were selected, a behavior.
typical scenario was set for transmission of digital content,
and a server was configured to respond to the user requests, From the characterization of the DTV load application
simulating an iDTV service infrastructure. Thus, other parameters that would be inserted in the simulation
characterization of the load of a typical digital TV was environment were set.
obtained. Figure 1 shows the infrastructure, with a
B. Definition of applications and parameters
transmission layer, responsible for transmitting the television
programs for broadcast and interactive applications. After characterization of load application, it was necessary
to conduct a research on commonly used flow on the Internet.
The idea is to make the environment of a user utilizing an
interactive application as close to a real scenario as possible.
So, three streams that were used in simulation tests were
defined: (i) application of video (video conferencing), (ii)
application of voice over IP (VoIP), (iii) typical application of
digital TV (DTV-Educ 2.0). These three applications aimed to
model traffic used by typical customers who use the Internet,
and the first two were taken from [21].
The probabilities of some configuration parameters within
Digital TV application were obtained (from empirical studies)
to characterize the traffic load and they were fed as input to
Fig. 1. Scenario tests the application of Digital TV the simulator. The specification of the traffic load for the
application DTV-Educ 2.0 is found in Table II.
The application is received and executed in the user layer,
which has a TV with a connected set-top box and access TABLE II. PARAMETERS FOR DIGITAL TV APPLICATION
technology for a return channel. The infrastructure of the Configuration parameter Value
return channel is represented in the service provider layer and HTTP Especification HTTP 1.1
can be wired or wireless, leaving it to the provider to decide Page Interarrival Time (seconds) Weibull (0.30419, 0.1139)
which technology is to be employed. Page Properties Lognormal (4.2996, 0.25489)
Type of Service Best Effort (0)
The layer of the test server is responsible for responding to
requests from tests, with only the application server, which has The configurable parameters for video traffic and VoIP
the function of helping in the measurement criteria used for application are presented in Table III and IV, respectively.
decision making.
TABLE III. PARAMETERS FOR VIDEO APPLICATION
The aforementioned scenario was set to extract
Configuration parameter Value
measurements of a typical network of Digital TV. The Frame Interarrival Time (sec) Constant (0.1)
application used in these tests was developed by the research Frame Size (bytes) Exponential(15625)
group of this article, and is called DTV-Educ 2.0 [20]. Type of Service Best Effort (0)
The purpose of this application, a chat conversation, is to
TABLE IV. PARAMETERS FOR VOICE APPLICATION
make it available for regular television program, so that a
student who is watching television can make classroom Configuration parameter Value
questions in real time or even conduct group work with other Silence Length (sec) Exponential(0.65)
students. Talk Spurt Length (sec) Exponential(0.352)
Encoder Scheme GSM (silence)
A. Characterization of load application Voice Frames per Packet 1
Type of Service Best effort (0)
To characterize the load application in the network, tests
were performed where two real users, for 10 minutes, Thus, applications have been developed and their
exchanged messages via chat provided by the application. characteristics were modeled using the parameters
configurable by the simulator.
All application information was recorded during the

155
 MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de
Eletromagnetismo

C. Simulation of a scenario In order to verify the application of regression, i.e., ability

The development of the simulation tests required a tool of Xi to explain variations in target variable Yi, multiple
that can simulate the performance of the network. Thus, coefficient of determination (R2) is:
OPNET Modeler was opted, which is widely used as a tool for
A t ( X t X ) A − nY 2
modeling telecommunications networks, and their work R2 =
environment allows creating a network from a library of yt y
templates, and setting parameters not only for the
environment, as well as for each object that is composed, and where Ῡ is the average value of Y, and y is the reduced
the impacts of its variations [22]. variable and centered around the average y=( Yi - Ῡ). It is
important, when using multiple linear regression model, to
Through this software, it was possible to observe the calculate the adjusted coefficient of determination according
behavior of a network based on WiMAX, with the parameters to:
defined in the previous section.
" n −1 %
Settings were made with parameters of real devices in R 2 = 1− (1− R 2 ) $ '
order to make the simulation as realistic as possible, by #n−k&
applying these settings in the simulator. In a similar way, the absence of dependencies (Pai=Ø) of
After analyzing the data input and configuration of the the variable in question is explained, for each possible
simulator, simulations were performed using a simulation time dependency relationship, when a value closer to zero or less
of 15 minutes to set up each scenario. This time was defined than zero is obtained for Ṝ2.
from literature of area. When applying multivariable linear regression, in order to
The tests were done thoroughly from the variation in the establish correlations between the variables in the domain, it
number of users on the network (until 40 users) and the was necessary to define the attributes (metrics) considered for
change of seed of the simulation, which meant that for the the analysis proposed in this paper. These metrics were
same amount of users, different values were obtained. An selected from the set of tests in the scenarios modeled in
extensive database was generated and consolidated, containing OPNET.
all the measured results of simulation experiments. Thus, among the metrics provided by OPNET for analysis,
D. Applying Multivariable Linear Regression seven were defined, which were used in the process of
decision making. These metrics are: Throughput, Jitter, Delay
After the organization of the database, a multivariable
Retransmissions, Object Response Time and Data Dropped.
linear regression was used to assess the influence (weight) that
each parameter has on the final result (selection of return From the definition of the metric (also called criteria) that
channel). would be used for decision making, Multivariable Linear
Regression was applied. To this end, an application was
The general formula of multiple regression models can be
developed.
specified as:
Multivariable Linear Regression, when applied on a set of
Yi = A0 + A1 X1i + A2 X 2i ++ Ak X ki + ui data, extracts the relationship of a given criteria with respect to
others. So, this relationship (A) will be used as a weight vector
A general multiple regression system may be represented
and it is the input to multicriteria models of decision making.
as matrix form:
The values of A were obtained and were normalized to
! Y $ ! 1 X X X $ ! A $ ! u $ yield the following:
11 12 1k
# 1 & # & # 0 & # 0 &
# Y2 & # 1 X 21 X 21  X 21 & # A1 & # u1 & A = [0,06 0,03 0,15 0,41 0,13 0,22]
# &=# &×# &+# &
#  & #     & #  & #  & 2
R (more than 90%) was calculated to verify the quality of
#" Yn &% #" 1 X n1 X n2  X nk &% #" Ak &% #" uk &% linear regression. The extracted weights are throughput, jitter,
delay, retransmissions, response time and discarded data.
Where: Y is column matrix of size n x 1 of observations or
records of dependent variable or target variable Y; X is a Thus, by extrapolating the fitted model and assuming that a
matrix of size n x k, i.e., n observations e k variables, in which service provider has more than one choice of access
the first column represents intercept A0; A is a column matrix technology for a given scenario, the purpose of this paper is to
of k x 1 unknown parameters; u is a column matrix with define what technology should be used.
perturbations.
IV. TESTS TO EVALUATE THE FIRST MILES
This specification (above equation) generates values to A
by employing Ordinary Least Squares - OLS. OLS estimates From the weights defined, real test scenarios and
parameters to A with least variance without bias and experiments were performed to test the proposed strategies.
minimizes the sum of squared errors of the estimated In order to perform measurements in real scenarios, three
regression. access technologies were used: PLC, WiMAX, ADSL and 3G.
The result of applying OLS to the presented model is given The choice of these different technologies are based on the
−1 availability of resources for testing and the focus of these tests
(
t
by: A = X X ) × X tY on heterogeneous networks since the metrics analyzed should
provide different values due to each technology’s peculiarities.

156
 MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de
Eletromagnetismo

WiMAX networks, for example, suffer interference caused

by external factors such as distance between antennas and TABLE VI. TOPSIS RESULT
similar frequencies, just like 3G. PLC networks, which use Alternative Ci
electricity to transfer data, suffer interference caused by ADSL2+ 0.99999
devices that operate in frequency bands similar to those used PLC 0.42772
for data transmission. ADSL technology uses fixed telephony WiMAX 0.30145
lines for data transmission and is characterized by impulsive 3G 0.26261
and background noises. Results shown in the Table refer to percentage of
Thus, scenarios were set for interactive digital TV indication to use a given return channel (option). So, ADSL2+
transmission in LPRAD (Laboratory of Network Planning for showed the best results followed by PLC, WiMAX and 3G.
High Performance), located at UFPA (Federal University of B. ELECTRE III Method
Pará), in order to test the different types of technologies, as
illustrated in Fig. 1, based on four layers already defined. In order to compare the results from TOPSIS, an algorithm
Then, a java applet to obtain the results of the criteria was developed to run ELECTRE III, analytical method. Tests
established was employed. were conducted with the same values used in Table V. The
decision matrix, Table VI, is the same used in the previous
Based on the sampling theory, the tests performed were method.
repeated over fifty times, to obtain an acceptable average for
comparison. The average results are shown in Table V. Relations of subordination were constructed to employ
ELECTRE III. Indices of particle agreement, global agreement
and credibility were determined according to the methodology
TABLE V. RESULTS WITH DIFFERENT TECHNOLOGIES of ELECTRE III.
PLC WiMAX ADSL 3G
Average Jitter (ms) 4,8 6,7 3,1 32,1 The second step was to classify the alternatives. Based on
Average Delay (ms) 16,6 21 13,5 135,00 the credibility indices, relations of subordinations. The final
Average Throughput (Kbps) 716 1436,6 6912,5 1804 stage was to construct a pre-ordering of intersection.
Object Response Time (s) 31,6 25,8 21,9 32,1 Intersection of two pre-ordering classifies more reliably
Number of Data Dropped 2.500 1.800 1.100 3000
proving that the options cannot be compared when they exist.
Number of Retransmissions 3.000 3.700 2.200 3200
After the intersection of all the alternatives, ranking was
The values shown in Table V show the average of the obtained as shown in Table VII.
results obtained from 1 to 40 users. These numbers represent
measurements from real-world scenarios at any given time and TABLE VII. RANKING OF TECHNOLOGIES WITH ELECTRE III
were used in the decision making process presented in this Alternative
paper. ADSL2+
WiMAX
From these values, a decision framework using TOPSIS PLC
and ELECTRE III was implemented to calculate decision 3G
making based on weights previously obtained.
Just as in the previous case, this also chose ADSL2+ as the
V. APPLYING MULTICRITERIA DECISION MAKING best followed by PLC, WiMAX e 3G. If, for some reason
After evaluating the networks, a decision of which (financial, political, availability), the first technology cannot
technology to be used follows. be chosen, the next one is recommended. It is noteworthy that
different analytical methods show a same result with slight
A. TOPSIS Method variations in percentage figures of indicating technologies.
For the decision making, TOPSIS method was applied on Therefore, any method may be employed for decision making.
the results measured from the network (Table V), based on the
vector A, which is the weight for each metric evaluated. VI. CONCLUSION
This paper presented a hybrid multicriteria model by
An application was used to execute the procedures of combining linear regression (to define weights for the criteria)
TOPSIS. Results show a relative closeness to the ideal and TOPSIS and ELECTREII decision making analytical
solution that ranked the options. The first matrix is the methods (to select the best access technology) to enable
decision matrix that consists of all the criteria and alternatives planning by service providers considering real scenarios
and their respective measured values (Matrix of table V). utilizing interactive digital TV of ISDB-T standard. So,
Criteria that should be maximized and minimized are also service providers may plan/predict the behavior of the services
defined. that are offered over the existing telecommunication
Once the normalized criteria matrix was obtained, weights infrastructure. Thus, they may avoid bottlenecks or idleness in
were assigned to criteria by multiplying the criteria by the a variety of scenarios.
assigned weights. A maximum value (ideal situation, positive) After careful and a thorough literature review, it was found
as well a minimum value (non-ideal situation, negative) was that the topics as well as open problems cover those discussed
determined for each evaluated item. Finally, approximate in this paper. It is important to point, however, that studies on
results (positive and negative) were obtained and the last combining multivariable linear regression with analytical
phase is to generate decreasing order of the obtained values as methods for decision making are in preliminary stage to be
shown in Table VI. applied in interactive digital TV environments.

157
 MOMAG 2014: 16º SBMO - Simpósio Brasileiro de Micro-ondas e Optoeletrônica e 11º CBMag - Congresso Brasileiro de
Eletromagnetismo

The paper discussed here combined the benefits of each [8] Hideyuki K. et al.. Combined Adaptive Congestion Control Method for
research to cope up with the main deficiencies mentioned in Communication-Broadcasting Integrated Services. In: Global
Telecommunications Conference. Annals… p. 1, 2006.
the literature, highlighting: (i) studies that use multicriteria [9] Hideyuki K. et al.. Access Control Method based on Sample Monitoring
analysis strategies that use analytical methods combined with for Volatile Traffic in Interactive TV Services. In: Global
multilinear linear regression; (ii) characterization of iDTV Telecommunications Conference. Annals… p. 1-6, 2008.
flow based on measurements obtained from actual scenarios; [10] Kabat MR, Patel MK, Tripathy CR. A heuristic algorithm for core
(iii) employing array of weights analyzed with linear selection in multicast routing. Journal Of Computer Science And
Technology 26(6): 954-961. Nov. 2011.
regression and not on specialists as suggested by other [11] Charilas, D. E. et al.. Application of Fuzzy AHP and ELECTRE to
methods, AHP for instance; (iv) use of different access Network Selection. Mobile Lighwieght Wireless Systems – Lecture
technologies already consolidated in the market. Notes of the Institute for Computer Sciences Social Informatics and
Telecommunications Engineering. V. 13, pp 63-73. 2009.
The main results obtained was the heuristic for [12] Sgora, A. et al.. An Access Network Selection Algorithm for
multicriteria decision that combined linear regression and Heterogeneous Wireless Environments. ISCC 2010.
TOPSIS and ELECTRE III analytical methods for decision [13] Sgora A. et al.. Network selection in a WiMAX-WiFi Environment.
Journal of pervasive and mobile computing. 2010.
making, considering IDTV. The proposed strategy is generic [14] CHEN, Y. Fuzzy AHP-based Method for Project Risk Assessment. In:
and flexible and can incorporate other parameters and CONFERENCE ON FUZZY SYSTEMS AND KNOWLEDGE
technologies that were not specified in this paper enabling new DISCOVERY (FSDK). Annals… Yantai, Shandong, p. 1249 – 1253.
decisions to different scenarios. 2010.
[15] YAN, G. et al.. The Design and Application of a Generic AHP
REFERENCES Evaluation System. In: 4th INTERNATIONAL CONFERENCE ON
WIRELESS COMMUNICATIONS, NETWORKING AND MOBILE
[1] Brazilian Institute of Geography and Statistics (IBGE). National Survey COMPUTING (WICOM). Annals… Donghua Univ., Shanghai, p. 1,
by Household Sample, 2009. Available in: 2008.
<[Link] [16] CHUNHAO, L. et al.. An improved ranking approach to AHP
o/pnad2009/[Link]>. Accessed: 31 May 2011. alternatives based on variable weights. In: 7th WORLD CONGRESS
[2] Xiao Y. et al.. Internet Protocol Television (IPTV): The Killer ON INTELLIGENT CONTROL AND AUTOMATION (WCICA).
Application For The Next-Generation Internet. IEEE Communications Annals… Jilin Univ., Changchun, p. 8255 - 8260, 2008.
Magazine. v. 45, e. 11, p. 126–134, 2007. [17] GELADI, P., KOWALSKI, B. R. Partial least-squares regression: a
[3] Dutta-roy, A. Bring home the Internet. Journal IEEE Spectrum. e. 12, v. tutorial. Analytica Chimica Acta, Amsterdan, v. 185, p. 1-17, 1986.
39, p. 32-38, 2002. [18] Sun Xuan ; Niu Qinzhou ; Xu Hefei ; A Bayesian Method for Decision
[4] Budri, A.K.; et al.. WiMAX Simulation Models for Return Channel in of Weight for MADM Model with Interval Data. Advanced Computer
Digital Television Systems. In: International Telecommunications Control, 2009. ICACC '09. International Conference on.
symposium. Annals... Fortaleza, v. 3, p. 688-693, 2006. [19] Decomposition of Variation in Experts´ Judgments in the Analytic
[5] Polo, E. I. et al.. PLC as a return channel for interactive digital TV. In: Hierarchy Process. 40th Conference on Computers and Industrial
International Conference on Communications and Networking in China. Engineering (CIE). 2010.
Annals… Shangai, China, p. 1-5, 2007. [20] SERUFFO, M. C. R.; MONTEIRO, F. P.; SILVA, C. N.; JACOB
[6] JAIME, G. D. G. et al.. User Level Fairness of a Multirate 3G JUNIOR, A. F. L.; LISBOA, Diego ; FRANCÊS, Carlos. DTV-EDUC
Technology (EVDO) With Application to Digital TV Return Channel 2.0: a case study of interactive educational application for Digital TV.
Technology. Broadcasting, IEEE Transactions on, 2008. In: 2o International Symposium on Digital Television, 2011.
[7] Nidal Nasser. A Theoretical Approach for Service Provider Decision in [21] CASTRO, M. C. Proposals for implementing quality of service in MPLS
Heterogeneous Wireless Networks. IEEE Communications Society VPN architecture using formal specification language SDL object-
subject matter experts for publication in the IEEE CCNC 2006 oriented and performance analysis using OPNET simulator. Dissertation
proceedings. submitted to the Faculty of Electrical Engineering and Computer
Sciences.
[22] OPNET Modeler. Extracted from [Link]

158