A study of traffic congestion analysis and monitoring system on internet of vehicles

To deal with this problem, this study proposes an MQTT-based traffic congestion analysis and monitoring system for the IoV,which includes data collection, segmented structure establishme

1! f :k

.<._hff- 1E

A Study of Traffic Congestion Analysis

and Monitoring System on Internet of

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

by

Nguyen Due Binh

A dissertation submitted to the graduate division in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

at Department of Information Engineering and Computer Science

Feng Chia University Taichung, Taiwan, R.O.C.

June 27, 2018

Approved by

)w

f}t

&t,J -f e., f(

First and foremost, I would like to express my greatest and sincere gratitude to

my advisor, Prof Chyi-Ren Dow for his motivation, immense knowledge, andpatience, especially for the continual supports of my Ph.D research It is an honorfor me to be accepted as one of his Ph.D students I am truly indebted to all hisideas, time and funding contributions that make my Ph.D experience energizing andproductive The passion and enthusiasm he has for the research were inspired andmotivational to me, especially during tough periods in the Ph.D pursuit I am alsograteful for the excellent advices he has offered as an outstanding professor Theseadvices encourage me in all the time of research and dissertation writing Fromthe bottom of my heart, I am sincerely indebted to him throughout my life time

In addition, I would like to thank my dissertation committee: Prof Shiow-FenHwang, Prof Tsan-Pin Wang, Prof Lien-Wu Chen, Prof Chun-Hua Chen, Prof.Meng- Yen Hsieh, and Prof Yi-Chung Chen, for their insightful encouragement, butalso for the meaningful suggestions which help me to widen my research fromvarious perspectives

I am indebted to Dr Jen-Cheng Chiu, Mr Po-Yu Lai, and Mr Kuan-ChiehWan, who provided me access to research materials and a great opportunity to work

in their team as an intern Without their precious support, it would not be possiblefor this research to be successfully conducted

I appreciate my mates at Mobile Computing lab for the inspiring discussions,unconditional supports, and for all the fun we have had in the last four years Inparticular, my grateful directly goes to Dr Yu-Hong Lee, who enlightens me the firstglance of research skills and Ms Yu-Yun Chang (Amber) for providing essential localsupports during the years My time at Feng Chia University and abroad was made asan

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

enjoyable period due to many friends became a part of my life I am appreciative for

my backpacking buddies and memorable trips to various famous places in Taiwan, fortime spent with roommates, for Hyde Lu’s hospitality as introducing me manyinteresting things about Taiwan’s culture, and for many other great friends I have evermet

Last but not least, I owe thanks to my family members for all theirencouragements and faith on me For my parents, who raised me up with a selflesslove and give me unlimited support for any decision I made For my brothers,who share the understanding and provide timely supports And most of all for myloving, supportive and patient wife and son whose faithful contribution during thefinal stages of this Ph.D is so appreciated

摘要近年來，車聯網已成為了一個興新的研究話題。該技術可以用來幫助現在的交通運輸系統更加智能化，並解決一些問題，如：交通壅塞。然而，現今在車聯 網環境下的運輸系統在解決交通壅塞問題時，無法滿足準確性、立即性以及相容 性的需求與挑戰。為了解決這個問題，本研究針對車聯網提出一個基於 MQTT 技術的交通壅塞分析與監控 系統系統，該系統包含了資料收集、建立分段結構、建立車 流模型、本地路段壅塞預測以及起迄路線壅塞預測等部分。這個系統從台中市政 府所提供的公開資料取得即時資訊，每個路段的交通流模型將會被正規化，並使 用基於模糊理論與路段標頭來預測本地路段的交通壅塞狀態。為了加強預測效率，這個研究也同時呈現了最小化錯誤率的驗證，該驗證基於 Rankine Hugoniot 條 件下。我們開發了一個使 用本方法提供給汽車的起迄交通估計服務，並實作了一 個雛形系統去驗證本方法的彈性，實驗結果也在精確度與系統回應時間上驗證了 此方法可以有效率地預測交通壅塞。

關鍵詞: 車輛偵測器, 消息隊列遙測傳輸(MQTT), 交通壅塞, 資料分析, 車流量

預測, 模糊理論, 起迄估計

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

Abstract

In recent years, the Internet of Vehicles (IoV) has been an emerging researchtopic Its techniques can be used to transform current transport systems intointelligent transport systems and resolve problems in transport, including trafficcongestion However, existing transport systems do not fully consider and resolveaccuracy, instantaneity, and compatibility challenges while resolving trafficcongestion in an IoV environment To deal with this problem, this study proposes

an MQTT-based traffic congestion analysis and monitoring system for the IoV,which includes data collection, segmented structure establishment, traffic-flowmodelling, local segment traffic congestion prediction, and origin-destinationtraffic congestion estimation The proposed system collects real data from opendata provided by Taichung City Government Macroscopic model-based traffic-flowfactors were formalized for each segment of the segmented structure on the basis ofthe analysis results obtained using the collected data Fuzzy rules-based localsegment traffic congestion prediction was performed using segment headers todetermine the traffic congestion state To enhance prediction efficiency, this studyalso presents a verification process for minimizing false predictions which is based

on the Rankine-Hugoniot condition An origin-destination traffic congestionestimation service for vehicles was developed on the basis of the proposedscheme To verify the feasibility of the proposed system, a prototype wasimplemented The experimental results demonstrate that the proposed scheme caneffectively predict traffic congestion in terms of accuracy and system response time

Keywords: Vehicle Detector Sensor, MQTT, Traffic Congestion, Data Analysis,

Traffic Forecasting, Fuzzy-based Rules, Origin-destination Estimation

Table of Contents

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

Acknowledgement i

摘要 iii

Abstract iv

Table of Contents v

List of Figures viii

List of Tables ix

Chapter 1 Introduction 1

1.1Motivation 2

1.2Overview of Research 7

1.3Thesis Organization 8

Chapter 2 Related Work 10

2.1Historical Traffic Data Analysis Strategies 10

2.2Mobility Structures for Data Sharing and Management in Vehicular Environment 12

2.3IoV and Light-weight Protocols 14

2.4Traffic Congestion Forecasting Methodologies 17

Chapter 3 Segmented Structure Scheme 21

3.1Segmented Structure Establishment 21

3.1.1VD Sensors Based Segment Demarcation 22

3.1.2E-Tag Sensors Based Segment Demarcation 23

3.2Traffic Data Collection Mechanism 25

3.2.1MQTT Publish/Subscribe Framework 25

3.2.2Vehicle Detector and E-tag Based Segment Data Collection 27

3.2.3Vehicular Data Collection 29

Chapter 4 Traffic Modelling and Analysis 31

4.1Traffic Flow Modelling 31

4.1.1Data Cleansing 32

4.1.2VD Based Segment Traffic Modeling 40

4.1.3E-tag Based Segment Traffic Modeling 41

4.2Traffic Congestion Observation Map 43

Chapter 5 Traffic Congestion Monitoring Scheme 50

5.1Local Segment Traffic Congestion Prediction 50

5.1.1Fuzzy-based Traffic Congestion Evaluation 50

5.1.2Traffic Congestion Condition Verification 52

5.2Origin-destination Traffic Congestion Estimation 57

Chapter 6 System Prototype and Implementation 62

6.1System Overview 62

6.2System Implementation 63

6.3System Prototype 66

Chapter 7 Experimental Results 69

7.1Analysis Results of Traffic Congestion Coefficient 69

7.2Average System Response Time 72

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

7.3Local Segment Traffic Congestion Prediction Results 75

7.4Origin-Destination Traffic Congestion Estimation Results 79

Chapter 8 Conclusions 81

8.1Summary 81

8.2Future Work 83

References 86

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

List of Figures

Figure 3.1 VD Sensors Based Demarcation 22

Figure 3.2 E-tag Sensors Based Demarcation 25

Figure 3.3 Publish/ Subscribe Framework 26

Figure 4.1 E-tag Segment Traffic Data Modelling 42

Figure 4.2 Traffic Congestion Observation Map 49

Figure 5.1 V e loci t y and Density Performance Index Membership Function 51

Figure 5.2 Traffic Flows MQTT Topic 55

Figure 5.3 Origin-destination MQTT Topic 59

Figure 5.4 Origin-destination Estimation 60

Figure 6.1 System Overview 63

Figure 6.2 Congestion Map on Control Center Interface 66

Figure 6.3 Section Information on Control Center Interface 67

Figure 6.4 Client Interface of Traffic Congestion Monitoring System 68

Figure 7.1 Traffic Congestion Coefficient Analysis Results of School Area 70

Figure 7.2 Traffic Congestion Coefficient Analysis Results of Business Area 71

Figure 7.3 Area-based Traffic Congestion Coefficient Comparison 72

Figure 7.4 Delay Time Comparison between MQTT and AMQP Protocols 73

Figure 7.5 System Response Time 75

Figure 7.6 Traffic Congestion Evaluation Results Using KNN Method 77

Figure 7.7 Traffic Congestion Evaluation Results Using the Proposed Method 78

Figure 7.8 Origin-destination Estimation Results 80

List of Tables

Table 3.1 Device Table 21

Table 3.2 VD Data Fields 28

Table 3.3 E-tag Data Fields 28

T a ble 3.4 R ea l - T ime V e h icle ’ s D a ta F ields 30

Table 4.1 Relative Rules 32

Table 4.2 Traffic Congestion Coefficient Table 49

Table 5.1 Fuzzy Rules for Traffic Congestion Evaluation 52

Table 5.2 Inflow Neighbor Table 55

Table 6.1 QoS Levels Assignment 64

Table 7.1 Parameters of SUMO Simulation 69

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

Chapter 1 Introduction

Transportation is one of none separable parts of human society and has closerelationship to other areas of society development Over the last few decades,numerous studies were proposed as the effort to make transportation become moreconvenient Various heterogeneous sensors and techniques [31, 65] were used totransform traditional transport systems into intelligent transport systems (ITS) andresolve problems in transport In recent years, traffic congestion has become a seriousproblem in cities, which not only negatively affects the daily lives of humans butalso impedes stable economic and societal development Traffic congestionincreases air pollution, travel time, and economic losses [13] These effects oftraffic congestion require analysis and forecasting scheme to monitor and reduce

in advance Governments increasingly strive to manage and resolve trafficcongestion; however, the task is difficult because of the complexity of theproblem; specifically, traffic congestion is difficult to predict Traffic congestionmay occur when slower cars share roads with faster cars, or when following carsmake a lane changing to overtake a bus exhibiting stop-and-go behavior, or anaccident suddenly on a road Generally, traffic congestion occurs when the number

of vehicles reaches the capacity of a road segment In this circumstance, theaverage vehicle speed and number of vehicles that pass the road segmentmarkedly decrease The complexity of traffic congestion is also reflected in itsdynamic and interrelated characteristics Traffic congestion can propagate from acongested road segment to neighboring road segments Because of these complexities,fully automatic analysis of traffic congestion is difficult to achieve Several methods[1, 4, 10, 48, 68, 72, 75] were applied to determine traffic congestion condition.These technologies are beneficiary in traffic congestion management; however, they

possess several drawbacks, such as, inaccurate traffic estimation, flexibilityproblem,

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

bandwidth problems and redundant data, especially in an urban environment

1.1 Motivation

ITS not only make traveling of individual vehicle become more convenient andenjoyable by providing different entertainment, information services and utilities, butalso provide a capability of traffic transition management [26], which is an essentialfactor for traffic planning, further road structure development, and resolve trafficproblems In the research field of traffic transition management to resolve trafficcongestion problem, the traffic flows information, including speed and direction ofvehicles are required However, the speed and directions of vehicles on the road aredifficult to predict because vehicles randomly move on road structure In the past,Vehicular Ad-hoc Networks (VANETs) based structures such as backbone structure[21, 27, 56, 64] were proposed to handle the requirement Because of vehiculartechnology shifting, the Internet of Vehicles (IoV) has been founded as the deeperintegration of the Internet of Things (IoT) to vehicular environment.Consequently, vehicles are equipped numerous types of sensors and techniques.Thus, additional mechanisms are required to handle the traffic data transmissions Asthe implementation, several studies [6, 20, 60, 73] investigate the capability ofpeculiar sensors in order to detect the traffic congestion condition such as camera,inductive loop detector, GPS devices to determine traffic congestion condition.However, these studies mainly focused on real-time traffic information services andlack of attention to the massive historical traffic data which can help to probe most

of traffic congestion situation in advance The reason involves the processingmassive historical data, a type of big data within the tolerable time in mobileenvironment is also a challenge On the other hand, as these characteristics isreflected in complexity of traffic congestion, the dynamic and interrelatedcharacteristics is also needed to be paid attention to handle traffic congestion problem

Fuzzy rules-based method is a particularly suitable solution for addressing complexnondeterministic problems, including traffic congestion predictions The existingsystems’ mechanisms simply classify traffic congestion levels by analyzing the trafficinformation on the basis of their own may result in false negative forecasting [76].Thus, a verification is important to guarantee the prediction results The major topic

of this study involves the traffic analysis based on historical traffic data, segmentedstructure establishment scheme, and segment-based traffic prediction in a largescale of road network scheme The motivation of each topic is described in detail asfollows

(1) Traffic data analysis

As the number of vehicles steadily grows, traffic congestion becomes one of themajor problems in cities According to statistical Taiwan’s preliminary data [101] thenumber of vehicles growth 186,776 during last three years The total number ofvehicle in Taiwan at the end of May 2017 was 21,585,949, in which cars areapproximately accounted for 37.5% In order to handle the traffic congestionproblem, traffic congestion patterns determination based on historical data is one offeasible solutions and it is normally considered as the first solution The patternsdetermination involves answering the question about time, location and duration thetraffic congestion usually occurs The reason is traffic congestions mostly recur atparticular road sections during particular time of at a location where the demandreaches or exceeds the capacity of road Determined patterns are able to help inreducing traffic congestion in certain roads as well as provide a better trafficmanagement Several studies [14, 18, 24, 52, 53] have proposed solutions forestimating short-term traffic congestion on the basis of time and location Thesesolutions based on a point of view, which can be comprehended as traffic

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

conditions at particular location, time and duration are closely correlated to the pastobservation results However, due to development of the IoV, various

heterogeneous sensors were integrated to ITS; the amount of producing data by thesesensors is incredible huge and grow extremely fast [49] These above-mentionedmethods cannot be deployed in the IoV environment Thus, this raises the need of amore suitable data analysis method, which can effectively handle a huge amount ofproduced data by both vehicles and sensors within the tolerable time in mobileenvironment to determine possible traffic congestion pattern at particular road sectionsduring particular time of the day and days of the week.

(2) Traffic congestion investigation approach

In the research field of traffic congestion problem, each segment of a roadnetwork exhibits different traffic-flow levels, and traffic congestion usually occurs

in high- traffic-flow road segments Traffic congestion starts at a specified roadsegment when the vehicle number reaches the capacity of the road segment.Thereafter, traffic congestion may propagate from a congested road segment toneighboring road segments Under this circumstance, every vehicle on the roadsegment will travel with the same pattern Because of this characteristic, it is required

to investigate traffic-flow levels of the road network based on traffic flowsinformation of individual road segments, which can be derived from the speed anddirections of vehicles on the road However, the speed and directions of vehicles aredifficult to predict because vehicles randomly travel on road structure Numerousstructures were proposed to investigate traffic-flow levels of the road network,which can be grouped into two categories, including centralized anddecentralized Some VANETs based studies utilized decentralized structures such

as the well-known grid structure [11, 21, 27, 59, 62, 64], which cooperates under thebackbone architecture to facilitate data management in a constrained area The keyfactor of these solutions is using headers to response for communication tasks with

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

nearby nodes in a grid Each header can also transmit data

with other headers to guarantee the transmission between vehicles over the wholenetwork However, due to the channel limitation and transmission protocol inVANETs [25], the data transmitted between headers may increases extra bandwidthconsumption and cause service delay On the other hand, several studies [23, 77, 79,83] proposed centralized structures, in which every vehicle will collect trafficinformation of surrounding environment, then send the collected information to adedicated server for traffic conditions evaluation Consequently, the evaluationresults can be sent back to vehicles as an assistance for vehicles’ traveling.Centralized structure may eliminate the above-mentioned disadvantage ofdecentralized structure; however, it may cause serious overhead problem when thevehicle number increases markedly in heavy traffic congestion situation Due tothese reasons, a more suitable structure is required to handle traffic congestionproblem.

(3) Traffic congestion prediction

Currently, two major challenges must be addressed to facilitate traffic congestionestimation and forecasting The first challenge is formalizing the basic factors fortraffic congestion management on a large-scale road network Several researchershave proposed solutions to resolve the problem, which can be grouped into twoapproaches, namely traffic congestion forecasting based on infrastructureequipments and traffic congestion forecasting based on Vehicular Ad-hoc Network(VANET) technology The first approach [23, 51, 77] mainly uses floating car data,namely data from Global Positioning System (GPS)-equipped vehicles Someprotocols use data from sensor equipment with the limitation, including loopdetectors, video recording devices, and smartphone-based technologies [19, 61, 67,96] Although the infrastructure-based approach is the most widespread and features

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

high reliability and accurate formalization, it lacks flexibility The second approach [5,

39, 66, 97] utilizes vehicle-to-vehicle (V2V)

and vehicle-to-infrastructure (V2I) technologies to gather vehicles’ traffic data, andthen formalizes the traffic model characteristics, including speed, density, flow, andtravel time Based on the formalized information, researchers have adoptedmathematical prediction algorithms to estimate traffic congestion levels Theadvantage to this approach is that it can be implemented without deployinginfrastructure sensors and has been demonstrated to operate effectively in varioustraffic and deployment scenarios However, the approach is limited by networkcommunication obstacles, including delayed and inaccurate traffic estimates [80],redundant data, bandwidth problems [87], and reliability problems [74] Theseproblems may cause insufficient precision or even failure in traffic congestionforecasting Moreover, the accuracy of the traffic modelling and traffic congestionprediction mechanisms mostly depends on the accuracy of the basic data gathering inthe processed area.

The second challenge is to guarantee the accuracy, instantaneity, and reliability oftraffic congestion estimation and forecasting Existing systems do not fully addressand resolve this challenge The VANET-based approach involves considerablepropagation delays and low reliability, whereas the infrastructure-based approachgenerally uses GPS data and data from limited number of sensors that lackflexibility Furthermore, most applications of the infrastructure-equipped approachuse HTTP-based protocols for gathering and transferring data between a centralcomputing unit and vehicles and equipment This not only hinders integration withvarious types of sensor equipment but may also cause a serious overhead problem ifthe vehicle number increases markedly in a traffic congestion situation or if theamount of data transferred over a long period of time increases substantially Bycontrast, because of the complexities of traffic congestion problems, these systems’mechanisms simply classify traffic congestion levels by analyzing the traffic

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

information on the basis of their own rules

without verifying the real-time traffic conditions This may result in false negativeforecasting [76].

1.2 Overview of Research

In consideration of aforementioned problems, this study aims to propose ananalysis strategy for historical traffic data to benefit traffic congestion evaluation and asegment-based traffic congestion analysis and monitoring system on Internet ofVehicle The overview of this study is described as follows:

a) First, the author proposes a segmented structure, which is converted fromgeographic map In the segmented structure, each segment is establishedbased on a vehicle detector (VD) device location or E-tag readers located onthe road One vehicle traveling on the road segment will be selected as theheader of the segment to responsible for data management and trafficcongestion evaluation within the segment For inter-communication,segments’ header is configured to communicate with the control center byusing a light weigh protocol in the IoV environment This structure has beendesigned to reduce communication redundancy and computationalbottlenecks as well as to consider sensors integration ability in IoVenvironment;

b) Second, the author proposes an analysis strategy for historical traffic data,including VD data and E-tag data The strategy is aimed to aid trafficcongestion evaluation on segments by determining segments with at a highrisk of traffic congestion during specific hours of the day and on specificdays of the week for further investigation This strategy helps to improveperformance of traffic congestion evaluation due to the priority ofevaluation is set on segments at a high risk of traffic congestion;

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

c) Third, to investigate traffic congestion patterns and its propagationphenomenon The author proposes a fuzzy rule-based mechanism andverification mechanism to evaluate traffic congestion condition for segments

In this scheme, the collected detector (VD) and E-tag data provided byTaichung City Government is used to formalize basic traffic flows factorsaccording to a macroscopic traffic-flow model Subsequently, the authordefines fuzzy rules according to Taichung City Government’s trafficcongestion classification framework As the mechanism’s design, the fuzzyrules-based traffic congestion evaluation is applied for high potential segment

of traffic congestion and performed by its header Moreover, the author alsodesigns a real-time origin-destination traffic congestion estimation module forvehicles on the basis of the fuzzy rule-based mechanism and verificationmechanism;

d) Finally, to verify the feasibility of system, the authors implement a systemprototype for both the control center and vehicles The system prototype is adistributed system that combines ubiquitous computing and shares trafficknowledge on the basis of MQTT The system implementation starts with theestablishment of MQTT protocol publish/subscribe methods Consequently,applications for the control center and vehicles are deployed to provide userinterfaces and communications

1.3 Thesis Organization

In summary, this research focuses on analysis strategy for historical traffic data tobenefit traffic congestion evaluation and a segment-based traffic congestion analysisand monitoring system on Internet of Vehicle The rest of this dissertation isorganized as follows Chapter 2 describes the current relevant research and

technology Chapter 3

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

discusses the establishment of segmented structure Chapter 4 discusses the datacollection method and proposed analysis strategy for historical traffic data in detail.Chapter 5 describes proposed segment-based traffic congestion forecasting mechanismand the proposed origin-destination traffic congestion estimation module for vehicles.Chapter 6 details the implementation prototype In Chapter 7, the experimental resultsare demonstrated Finally, Chapter 8 presents conclusions and future researchdirections

Chapter 2 Related Work

In this chapter, the author presents general reviews about the related work inresearch field of traffic congestion analysis and forecasting in IoV environments.Numerous studies have been discussed and proposed in the field This dissertationfocuses on historical traffic data analysis strategies, mobility structures for datasharing and management in vehicular environment, IoV and light-weight protocols,and traffic congestion forecasting methodologies The above-mentioned topics aredescribed as follows

2.1 Historical Traffic Data Analysis Strategies

Historical data play a significant role in prediction Especially, in trafficcongestion prediction, historical traffic data can help to effectively determine trafficcongestion patterns The historical data-based traffic congestion determinationconcerns questions about the time, location and duration the traffic congestion usuallyoccurs and normally considered as the first solution in traffic congestion prediction

To answer these questions, a historical traffic data analysis will be performed Thiscan be done because of a characteristics of traffic congestion that traffic congestionsmostly tend to recur during particular time at particular road sections, where thedemand reaches or exceeds the capacity of road Several studied has beeninvestigated in this field since 1980’s Ahmed and Cook [2] presented an statisticalmethod based on extracted from historical data to predict future traffic conditions,namely Auto-Regressive Integrated Moving Average (ARIMA), which wasconsidered as the earliest work in this field ARIMA is a time series data method andbasically based on three terms, including autoregressive term, non-seasonaldifferences, and lagged forecast errors Various later methods have been found as

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

special cases of ARIMA such as Random Walk [70], KARIMA [89],

seasonal ARIMA (SARIMA) [92], space time AIRMA (STARIMA) [52] Thesestatistical methods relied on solid mathematical foundations and mostly effective withthe availability of a prior knowledge about the functional relationship between trafficvariables However, constraining assumptions of these methods may fail when facingthe nonlinear and high complex data from heterogeneous domains To overcome theshortcoming of data gathering from heterogeneous domains, Furtlehner et al [35]presented a set of methods, which aim to extract large scale features of road traffic,both spatial and temporal The features are based on computed local traffic indexesfrom collected data of fixed sensors and floating car, which is artificially generatedaccording to mesoscopic model These methods rely on traditional data miningtechniques such as clustering and statistical analysis As the results, the authorsidentified traffic congestion patterns on a large scale of road network by usingbelief-propagation (BP) algorithm and the approximate Markov random field (MRF).

On the other hand, as an improvement based on traditional methods, Lin et al [63]proposed a novel floating car data analysis method based on data cube for congestionpattern exploration The method proposed by Lin is different from traditionalmethods, according to the numerical statistics of traffic data and multi-dimensionalanalysis framework In this method, a spatial-temporal related relationship on aslow-speed street is used to identify traffic congestion event Consequently, thetraffic congestion event is aggregated by a cluster style to determine the trafficpattern based on spatial-temporal dimension levels The congestion pattern can berepresented through aggregated location, time and duration and importantcongestions The experimental results show that the method can effectively identifyand summarize the congestion pattern in terms of computation and storage cost based

on the using of a historical traffic dataset collected during one week on a large urbanarea by more than 12000 taxis

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

2.2 Mobility Structures for Data Sharing and Management in

Vehicular Environment

In vehicular environment, it is required to divide a large-scale road network tosub-parts or clusters for easier data sharing and management This strategy not onlybenefit vehicles’ communication but also help to deal with traffic problem, includingtraffic congestion problem Numerous studies have proposed structures, which can begrouped into centralized and decentralized approaches Most of VANETs basedstudies utilized decentralized structures Dow, et al [28] proposed an efficientscheme to discover k services in cluster-based mobile ad hoc networks, calledAnyKast In the proposed scheme, an anycast tree based on the clustering and virtualbackbone is used to reduce unnecessary transmission Moreover, scope flooding isused to limit the information transmission piggybacking Subsequently, a periodicalinquiry mechanism is used to increase the service information accuracy As the result,the proposed schemes can effectively discover services, reduce requests, lowerscontrol overhead and searching latency Tsai, et al [88] proposed a grid-based servicediscovery protocol for Ad-hoc networks In the proposed scheme, information service

is registered according to predefined trajectory in a 2D logical grid architecture.Consequently, requestors can discover services along the trajectory and acquiresuitable services’ information The authors also provided an improved process toavoid the sparse node network topology The experimental results show that theproposed protocol outperforms than others protocol in terms of discovery successratio and the discovery cost In [27], Dow, et al presented schemes to effectivelydisseminate and discover service information based on grid architecture With theaid of public transportation infrastructure, the broadcast storm problem has beeneliminated using backbone structure The backbone structure

is created by bus routes and used to post and circulate data Experimental results showthat the proposed scheme archived higher packet delivery ratio, reduces end-to-enddelay and lowers overhead than existing schemes Chen, et al [15] proposed a grid-based routing protocol for VANETs, which uses map data to effectively generate

a shortest transmission grid route within tolerable computation time HarpiaGrid aims

to reduce unnecessary transmissions by restricting packets in grid sequences.Moreover, a local recovery scheme also presented and provided superior fault-tolerance capability As an improvement of grid based and virtual backbonestructure, Lee, et al [58] proposed a geo–aware tree–based service–trackingscheme to locate a specific service in a particular area The virtual backbone ispredefined according to the main roads in the city Consequently, geo–grids is furtherorganized into a tree structure, which is the higher layer of the virtual backboneand can be used to facilitate service location tracking This method makes nodeseasier to find a specific service and access data of interest in VANETs Theexperimental results indicated that the proposed approach outperforms existingschemes regarding the tracking success rate and service–tracking time In [30], Dow,

et al applied grid based architecture with geo-ware information system to benefittaxi carrying management system on Internet of things The proposed system allowsdrivers to both hunt on the road and wait in a queuing zone To enhance serviceefficiency and quality, a scheme is provided to prevent the ping-pong effect which

is based on the location-based services The results show the proposed schemeoutperforms the waiting and hunting models in terms of number of customers,vacancy rate, and profit The key factor of these studies is using a header toresponse for communication tasks of each grid Each header can alsocommunicate with nearby headers to guarantee the transmission of overallnetwork However, this approach persists limitations as channel limitation and

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

transmission protocol in VANETs, the data transmitted between headers mayincreases extra bandwidth consumption and cause

service delay.

On the other hand, several studies utilized centralized structures, in which eachvehicle may collect traffic information of surrounding environment through sensorsand directly communicate with a dedicated server The dedicated server response fordata sharing and management of overall network Sagar and Shilpa [77] proposed acontext-aware traffic congestion management system to discover the current status ofeach vehicle and the density of the road Based on the discovered information, systemdynamically adjusts traffic signals according to the surrounding environmentalconditions The system implementation includes hardware module and softwaremodule and is deployed for both server side and vehicle side The system is able tocollect real-time collection, organization to provide an efficient and accurateestimation of traffic density and weather condition Stojanović, et al [84] presented aWeb-based information system for fleet management purpose, called MOWIS.MOWIS is a service-oriented open software was used to develop the system based

on platform for mobile and Web applications The proposed system used based and context- aware information to provide services for vehicles, includingdirectory service, routing service, traffic information service and real data ofautomatic vehicle location bus tracking service A Web server has been developed topossess all functionalities needed for these features In general, centralized structuremay eliminate the mentioned disadvantages of decentralized structure; however,centralized approach mostly used traditional HTTP-based protocols forcommunications between fleet control center and vehicles which may cause seriousoverhead problem when the vehicle number increases markedly in traffic congestionsituations

location-2.3 IoV and Light-weight Protocols

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

The Internet of Things (IoT) era has opened up as an emerged research trend over

recent years The IoT has originated the needs of better monitoring, control andmanagement in various areas, such as healthcare [8], education [22], entertainment[45], and transportation [69] The IoT was first mentioned by Gates, et al [36], whichcan be defined as an interconnected network of things, including humans, vehicles,machines, sensors, etc., which have the ability to measure and act all over theworld by using communication technologies The key factor of the IoT is to obtaininformation about surrounding environment to understand, act and control on it TheIoV plays as a part of the IoT, but it has distinctive characteristics In the IoV, themobility of vehicles is an important topic that needs to be paid attention The IoVinvolves the way to gather information regarding vehicles, roads, and theirsurroundings Moreover, the IoV also requires the processing, computing, and securetransmission information onto mobility platforms with the integration of theInternet Based on collected information, IoV systems can effectively supportvehicles by providing numerous services, including entertaining, guiding,surveillance, protecting from unattended collision IoV technologies aim to use inintelligent transportation systems (ITS) which deliver intelligent traffic management,intelligent dynamic information services, and intelligent vehicle control as thetypical IoT applications in mobility environment In recent years, the IoV isproposed to provide more convenient services Yang, et al [94] proposed anabstract IoV network model, which considers various connections of vehicles,roads, environments and pedestrians Singh and Singh [78] specified connectedvehicle architecture solutions for both personal and public vehicles to providesmart driving and improve safety for the vehicles The main idea behind this work is

to utilize the dashboard camera on vehicles to enhance the control and accidentprevention/monitoring services The dashboard camera was used to capture and sharevehicle’s real-time accident/traffic footage to the related authorities such as nearest

A Study of Traffic Congestion Analysis and Monitoring System on Internet of Vehicles

vehicles, police staff, hospital, family members and insurance company instantlyalong