Introduction
Motivation
ITS enhances individual vehicle travel by offering entertainment, information services, and utilities, making journeys more convenient and enjoyable Additionally, it provides essential traffic transition management capabilities, crucial for traffic planning, infrastructure development, and congestion resolution Effective traffic transition management relies on accurate traffic flow information, including vehicle speed and direction; however, predicting these parameters is challenging due to the random movement of vehicles on road networks.
Vehicular Ad-hoc Networks (VANETs) based structures such as backbone structure [21,
The Internet of Vehicles (IoV) has emerged as a result of advancing vehicular technology and represents a deeper integration of the Internet of Things (IoT) within vehicular environments Vehicles are now equipped with numerous sensors and communication techniques, necessitating additional mechanisms to efficiently manage traffic data transmissions Several studies have been proposed to address these challenges and optimize data handling in IoV systems.
Recent studies have explored various sensors such as cameras, inductive loop detectors, and GPS devices to detect traffic congestion effectively However, these approaches primarily focus on real-time traffic information services and often overlook the potential of large-scale historical traffic data for predictive analysis Processing massive volumes of historical data within a mobile environment presents significant challenges due to computational complexity and data volume Additionally, the complex, dynamic, and interconnected nature of traffic congestion requires attention to the evolving relationships and patterns within traffic data to develop more accurate and proactive congestion management strategies.
This document is for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Fuzzy rules-based methods are highly effective for addressing complex, nondeterministic problems such as traffic congestion prediction Existing systems often classify congestion levels by analyzing traffic data, but this approach can lead to false negatives without proper verification, highlighting the need for reliable validation methods This study focuses on traffic analysis using historical traffic data, developing segmentation schemes, and implementing segment-based traffic prediction across large-scale road networks The motivation behind each of these topics is thoroughly discussed to improve the accuracy and reliability of traffic forecasting systems.
As urban vehicle numbers continue to rise, traffic congestion has become a major city issue, with Taiwan experiencing a vehicle growth of 186,776 over three years, totaling nearly 21.6 million vehicles by May 2017, of which 37.5% are cars To address this challenge, analyzing traffic congestion patterns based on historical data is a primary and effective solution, as it helps identify when, where, and how long congestion typically occurs Congestion often recurs at specific road segments during peak times when demand exceeds capacity, and understanding these patterns facilitates targeted traffic management strategies Numerous studies have proposed short-term congestion estimation methods focusing on location and time, leveraging past observations to predict traffic conditions However, advancements in the Internet of Vehicles (IoV) are enhancing these approaches, enabling more dynamic and accurate traffic management solutions.
This document is for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, as well as authorization from the author and the author's thesis advisor.
Integrating heterogeneous sensors into Intelligent Transportation Systems (ITS) has led to an exponential increase in data generation, posing significant challenges for data analysis Existing methods are unsuitable for the IoV environment due to the vast volume and rapid growth of data from both vehicles and sensors Therefore, there is a critical need for advanced data analysis techniques capable of efficiently processing large-scale data within acceptable timeframes Such methods are essential for accurately identifying traffic congestion patterns on specific road sections at particular times of the day and days of the week.
In traffic congestion research, different road segments exhibit varying traffic flow levels, with congestion typically occurring in high-traffic areas Congestion begins when the number of vehicles on a segment reaches its capacity and can spread to neighboring segments During congestion, all vehicles on the affected road segment tend to follow a uniform travel pattern Therefore, analyzing traffic-flow levels across the entire network requires detailed traffic flow data from individual segments, which can be obtained by monitoring vehicle speeds and directions.
Predicting vehicle speed and direction remains challenging due to their random movement across road structures To analyze traffic flow in road networks, various models have been proposed, broadly categorized into centralized and decentralized structures Many VANET studies favor decentralized approaches, such as grid-based structures, which operate under a backbone architecture to manage data within confined areas These solutions primarily rely on headers for communication with nearby nodes in the grid, enabling efficient data transmission and network management.
This document is intended solely for viewing purposes; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
5 FCU e-Theses & Dissertations (2018) with other headers to guarantee the transmission between vehicles over the whole network However, due to the channel limitation and transmission protocol in VANETs
Transmiting data between headers can increase bandwidth consumption and cause service delays Several studies have proposed centralized traffic management structures where vehicles send environmental data to a dedicated server for traffic assessment, and the evaluated results are communicated back to assist vehicle movement While centralized architectures can overcome some decentralized limitations, they may lead to significant overhead issues in high-traffic scenarios with numerous vehicles Therefore, developing a more efficient and scalable traffic management structure is essential to effectively address congestion challenges.
Effective traffic congestion estimation and forecasting face two primary challenges: formalizing key factors for managing congestion on large-scale road networks, and developing accurate prediction methods Researchers have proposed two main approaches—one relies on infrastructure-based data collection, utilizing floating car data from sources like GPS and sensor networks, while the other leverages Vehicular Ad-hoc Network (VANET) technology to gather real-time vehicle communication data Addressing these challenges is essential for enhancing traffic management systems and reducing congestion efficiently.
GPS-equipped vehicles and sensor-based protocols, such as loop detectors, video recording devices, and smartphone technologies, are commonly used for traffic data collection, although they face limitations in flexibility Infrastructure-based traffic management methods are widespread due to their high reliability and precise formalization, but they lack adaptability to dynamic traffic conditions Alternatively, vehicle-to-vehicle (V2V) communication approaches offer enhanced flexibility by enabling direct data exchange between vehicles, improving traffic flow and safety.
This document is for viewing purposes only; any other use requires prior consent from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Overview of Research
This study addresses the challenges of analyzing historical traffic data by proposing an effective analysis strategy to enhance traffic congestion evaluation It also introduces a segment-based traffic congestion analysis and monitoring system tailored for the Internet of Vehicles, aiming to improve real-time traffic management and reduce congestion issues.
This study introduces a segmented structure derived from geographic maps, where each segment is defined based on the location of vehicle detector (VD) devices or E-tag readers installed along the road This innovative approach enables more precise traffic monitoring and management by leveraging the strategic placement of detection devices.
In this traffic management system, a designated vehicle within each road segment acts as the header responsible for data collection and evaluating traffic congestion These segment headers communicate with the central control center using a lightweight protocol, ensuring efficient inter-vehicle and infrastructure communication This approach enhances real-time traffic monitoring and enables effective congestion management across road segments.
The IoV environment is designed to reduce communication redundancy and computational bottlenecks while enhancing sensor integration capabilities Additionally, an analysis strategy for historical traffic data, including VD and E-tag data, is proposed to evaluate traffic congestion on specific segments This strategy focuses on identifying segments at high risk of congestion during particular hours and days, allowing for targeted investigations and improved traffic management By prioritizing high-risk segments, this approach effectively enhances the accuracy and efficiency of traffic congestion evaluation.
This document is intended solely for viewing purposes; any other use requires prior authorization from the National Central Library of Taiwan, the author, and the author's thesis advisor.
In 2018, 8 FCU e-Theses & Dissertations highlighted the importance of analyzing traffic congestion patterns and their propagation phenomena The study introduced a fuzzy rule-based mechanism designed to assess traffic congestion conditions across different segments Additionally, a verification mechanism was proposed to ensure accurate evaluation of congestion levels, contributing to more effective traffic management solutions.
In this scheme, the collected detector (VD) and E-tag data provided by
The Taichung City Government utilizes a macroscopic traffic-flow model to formalize key traffic flow factors Building on this, fuzzy rules are defined in accordance with the city's traffic congestion classification framework to evaluate congestion levels A fuzzy rule-based mechanism identifies high-potential traffic congestion segments and assesses their severity Additionally, a real-time origin-destination traffic congestion estimation module is designed to monitor vehicle flow effectively To verify system feasibility, a distributed prototype integrating ubiquitous computing and MQTT protocol is implemented for both control centers and vehicles, facilitating traffic data sharing and user interfaces.
Thesis Organization
This research focuses on analyzing historical traffic data to enhance traffic congestion evaluation and develop a segment-based traffic congestion analysis and monitoring system within the Internet of Vehicles The study aims to improve traffic management by leveraging data-driven insights The dissertation is structured as follows: Chapter 2 reviews existing research and technological advancements relevant to traffic analysis, while Chapter 3 explores methodologies and system design for traffic congestion monitoring.
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
9 FCU e-Theses & Dissertations (2018) discusses the establishment of segmented structure Chapter 4 discusses the data collection method and proposed analysis strategy for historical traffic data in detail
Chapter 5 describes proposed segment-based traffic congestion forecasting mechanism and the proposed origin-destination traffic congestion estimation module for vehicles
Chapter 6 details the implementation prototype In Chapter 7, the experimental results are demonstrated Finally, Chapter 8 presents conclusions and future research directions
This document is for viewing purposes only Any additional use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Related Work
Historical Traffic Data Analysis Strategies
Historical traffic data is essential for predicting traffic congestion, as it helps identify recurring patterns related to time, location, and duration of congestion Analyzing past traffic data provides insights into when and where congestion typically occurs, especially since traffic jams often recur during specific periods on certain road sections when demand exceeds capacity This approach is considered the primary method in traffic congestion prediction Numerous studies have explored this field since the 1980s, emphasizing the importance of historical data in understanding and forecasting traffic conditions.
Ahmed and Cook [2] presented an statistical method based on extracted from historical data to predict future traffic conditions, namely Auto-Regressive Integrated Moving
ARIMA (AutoRegressive Integrated Moving Average) is one of the earliest and most foundational methods for analyzing time series data, relying on three key components: autoregressive terms, non-seasonal differences, and lagged forecast errors Over time, several advanced techniques, such as Random Walk and KARIMA, have been developed as special cases of ARIMA, expanding the range of models available for time series forecasting and analysis.
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Traditional statistical methods like seasonal ARIMA (SARIMA) and space-time ARIMA (STARIMA), grounded in solid mathematical principles, are effective when prior knowledge of traffic variable relationships is available; however, their assumptions often fail with complex, nonlinear data from heterogeneous sources To address this, Furtlehner et al introduced methods that extract large-scale spatial and temporal traffic features using local traffic indexes derived from fixed sensors and floating car data, employing data mining techniques such as clustering and belief propagation to identify congestion patterns across extensive road networks Building on traditional approaches, Lin et al proposed a novel data cube-based analysis for floating car data, leveraging multi-dimensional spatial-temporal relationships to detect congestion events on slow-speed streets, summarizing these patterns through location, time, and duration metrics Experimental results demonstrate that this method effectively identifies and summarizes congestion patterns while maintaining low computational and storage costs, utilizing a week's worth of taxi-derived traffic data from a large urban area.
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, the author, and the thesis advisor.
Mobility Structures for Data Sharing and Management in Vehicular
In vehicular environments, dividing large-scale road networks into sub-parts or clusters simplifies data sharing, enhances vehicle communication, and addresses traffic congestion issues Studies mainly focus on decentralized structures for VANETs, with protocols like Dow et al.'s AnyKast scheme efficiently discovering services using cluster-based ad hoc networks, reducing unnecessary transmissions through an anycast tree and scope flooding Additionally, periodic inquiry mechanisms improve service information accuracy, resulting in decreased request overhead, lower control costs, and reduced search latency.
In ad-hoc networks, the proposed scheme registers information services based on a predefined trajectory within a 2D logical grid architecture, enabling requestors to efficiently discover services along this path This approach allows users to acquire relevant service information dynamically as they move through the network To address challenges associated with sparse node network topologies, the authors have also introduced an improved process that enhances network connectivity and service availability across the ad-hoc network.
The experimental results demonstrate that the proposed protocol outperforms other protocols in terms of discovery success ratio and discovery cost, highlighting its efficiency and effectiveness According to Dow et al [27], innovative schemes have been developed to effectively disseminate and discover service information within grid architectures By leveraging public transportation infrastructure, these schemes eliminate the broadcast storm problem through the implementation of a backbone structure, ensuring reliable and scalable service discovery.
This document is intended for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, as well as the consent of the author and the author's thesis advisor.
The 13 FCU e-Theses & Dissertations (2018) highlights the use of bus routes for data posting and circulation in transportation networks Experimental results demonstrate that the proposed routing scheme achieves higher packet delivery ratios, reduces end-to-end delays, and lowers network overhead compared to existing methods Chen et al [15] introduced a grid-based routing protocol for VANETs that utilizes map data to efficiently generate the shortest transmission routes within acceptable computation times HarpiaGrid further enhances network performance by minimizing unnecessary transmissions through restricted packet routing in grid sequences, and its local recovery scheme offers improved fault-tolerance and robustness in vehicular networks.
Lee et al proposed a geo-aware tree-based service tracking scheme that enhances grid-based and virtual backbone structures for efficient service location within specific areas, leveraging predefined virtual backbones aligned with city roads By organizing geogrids into a hierarchical tree structure, this method simplifies node access to targeted services and data in VANETs, resulting in higher tracking success rates and reduced service-tracking times Similarly, Dow et al applied a grid-based architecture combined with geo-information systems to optimize taxi management in IoT environments, enabling drivers to hunt and wait efficiently while preventing ping-pong effects through location-based schemes These approaches utilize headers for communication within each grid and between neighboring headers to ensure reliable network transmission; however, they face challenges such as increased bandwidth consumption and transmission protocol limitations due to VANET channel constraints.
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
14 FCU e-Theses & Dissertations (2018) service delay.
Several studies have utilized centralized traffic management systems, where each vehicle collects environmental data via sensors and communicates directly with a dedicated server responsible for data sharing and network management For example, Sagar and Shilpa [77] developed a context-aware traffic congestion management system that monitors vehicle status and road density, dynamically adjusting traffic signals based on real-time environmental conditions This system includes both hardware and software modules deployed on servers and vehicles, enabling accurate traffic density and weather condition estimations Similarly, Stojanović et al [84] introduced MOWIS, a web-based fleet management system using a platform that integrates location-based and context-aware information to provide services such as routing, traffic updates, and automatic vehicle location tracking While centralized structures can mitigate some challenges of decentralized systems, they often rely on traditional HTTP protocols, which may generate significant overhead during high traffic congestion with many vehicles.
IoV and Light-weight Protocols
The Internet of Things (IoT) era has opened up as an emerged research trend over
This document is solely for viewing purposes; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Recent research highlights the growing importance of IoT in various sectors, with the Internet of Vehicles (IoV) emerging as a vital subset focusing on vehicle mobility, safety, and intelligent transportation systems The IoT, first introduced by Gates et al., is defined as an interconnected network of humans, vehicles, sensors, and machines capable of measuring, processing, and acting across the globe via communication technologies A key aspect of IoV involves gathering and securely transmitting information about vehicles, roads, and their surroundings to support services like traffic management, accident prevention, and real-time monitoring Advanced IoV systems utilize vehicle sensors, such as dashboard cameras, to enhance safety by capturing and sharing real-time traffic footage with relevant authorities, thus improving accident response and prevention Recent models consider diverse connections among vehicles, infrastructure, and pedestrians, aiming to deliver smarter, safer, and more efficient transportation solutions through intelligent traffic management and vehicle control.
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the thesis advisor.
In 2018, 16 FCU e-Theses & Dissertations highlighted innovative solutions for urban challenges, with Fu et al proposing an effective reservation optimization parking recommendation model that actively guides drivers based on evaluated indicators This phased approach improves parking efficiency by combining real-time data interaction with driver recommendations Additionally, Handte et al introduced advancements in urban transportation through their research on urban bus systems, contributing to more sustainable and efficient city transit solutions.
Navigator (UBN) is an innovative IoT-enabled navigation system designed specifically for urban bus riders, integrating IoT technology into public transport to enhance the user experience By leveraging IoT paradigms, it aims to eliminate barriers to public transportation usage, making bus journeys more accessible and convenient This advanced system not only improves navigation but also fosters a positive perception of public transit, encouraging more people to choose buses as their preferred mode of transportation.
Cheng et al [17] review routing protocols in the Internet of Vehicles, focusing on routing algorithms and their evaluation methods Their study categorizes these protocols into five taxonomy types: transmission strategies, information requirements, delay factors, applicability across different dimensions, and target networks This comprehensive classification aids in understanding the diverse routing solutions tailored for Intelligent Transportation Systems.
As the Internet of Vehicles (IoV) continues to expand, connecting numerous sensors and vehicles to the internet poses significant challenges for network bandwidth Developing lightweight communication protocols tailored for IoT and IoV environments is essential to enhance mobile network efficiency MQTT emerges as a leading candidate due to its publish/subscribe model, which reduces unnecessary communication, conserves energy, and lowers latency compared to traditional HTTP protocols Studies by Yokotani and Sasaki [95], along with Caro et al [12], confirm that MQTT offers superior scalability and reliability for high-frequency data transmission required in IoV applications The key benefit of MQTT’s publish/subscribe approach is its ability to minimize unnecessary connections between senders and receivers, optimizing network performance in vehicle communication networks.
This document is intended for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the thesis advisor.
In 2018, 17 FCU e-Theses & Dissertations highlighted the challenges of unstable network topology, prompting Lee et al [57] to conduct an in-depth comparison of the advantages across three QoS levels in the MQTT protocol Their findings demonstrate that managing QoS levels with varying payload sizes can significantly enhance network efficiency Additionally, Campo et al [9] proposed an integrated approach utilizing different degrees of QoS to optimize IoT communication performance.
The MQTT protocol architecture enables effective and reliable distribution of notification messages among different actors, making it ideal for applications such as home monitoring for patients with dementia Naik [71] provides an overview of four established IoT protocols—CoAP, MQTT, AMQP, and HTTP—highlighting their characteristics and offering a comprehensive comparison based on relevant criteria This comparative analysis helps users understand the strengths and limitations of each protocol, guiding them to select the most suitable one for their specific IoT system requirements Additionally, Szabó and Farkas [85] demonstrate the use of MQTT’s publish/subscribe model for designing smart city applications, showcasing its effectiveness in facilitating seamless communication in complex urban environments.
The design aims to apply to crowd-sourcing based smart applications, such as smart travel planner or smart parking in the cities.
Traffic Congestion Forecasting Methodologies
Traffic congestion has significant negative impacts on people's lives, including increased transportation costs, reduced productivity, and worsened pollution Numerous studies have highlighted these issues, emphasizing the urgent need for effective solutions to alleviate traffic jams and improve urban mobility.
Traffic congestion is a universal issue affecting both developing and developed countries worldwide [38] In transportation, traffic congestion is generally classified into two categories: recurrent and non-recurrent congestion [46], with recurrent congestion occurring regularly during peak hours.
This document is intended for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Traffic congestion is primarily caused by factors such as increased traffic flow during peak hours and the frequent on-ramp and off-ramp activities, with non-recurrent congestion resulting from road accidents, natural disasters, construction zones, or special events To address these issues, various traffic management strategies have been developed, utilizing advanced traffic flow detection techniques Hsiao et al introduced a vision-based real-time monitoring system that processes video frames from stationary cameras to detect vehicles and backgrounds, enabling efficient traffic event analysis Gholve and Chougule proposed a wireless sensor network system using magnetic sensors and an Arduino-based protocol for highway congestion detection Bhoraskar et al leveraged smartphone sensors combined with non-intrusive methods to monitor traffic conditions in Mumbai, with layered system architecture providing users with real-time traffic information Faye et al deployed wireless sensor networks at intersections, employing a distributed algorithm called TAPIOCA to enhance traffic congestion detection.
This document is for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
The 19 FCU e-Theses & Dissertations (2018) introduced the AdaPtive IntersectiOns Control Algorithm, which dynamically determines traffic light signal sequences to reduce average waiting times while minimizing starvation risk, demonstrated through SUMO simulation results showing improved queue size, throughput, and overall traffic flow Zhou [99] proposed a real-time traffic-responsive strategy that adaptively optimizes traffic signal timings based on current conditions, with simulation results indicating it outperforms other methods in reducing average waiting times Additionally, Winkler et al [93] explored congestion pricing as an effective solution to traffic congestion, developing a comprehensive framework for evaluating various congestion pricing plans to manage traffic demand efficiently.
Consequently, the optimal tolling theories are applied to sample small networks and the
The Sioux Falls test network experiment highlights the significance of tolling location and levels in traffic management Souza et al [81] introduced SPARTAN, a distributed intelligent transportation system (ITS) solution designed to mitigate traffic congestion by re-routing vehicles SPARTAN effectively informs drivers about congested areas, enhancing traffic flow and reducing congestion in urban networks.
SPARTAN utilizes V2V techniques and real-time decision-making to suggest re-routed paths for vehicles, effectively avoiding congested areas and reducing travel and congestion times in urban environments Simulation results demonstrate that SPARTAN outperforms existing approaches in managing traffic flow Additionally, Taneva and Davcev [86] proposed leveraging crowdsourcing and social networks, particularly utilizing Instagram data, as innovative solutions to address traffic congestion challenges.
HTTP-based methods for gathering and disseminating urban traffic information demonstrate significant potential in addressing traffic congestion issues within smart cities Early results indicate that these applications can effectively improve traffic flow and enhance urban mobility, making them a vital component of sustainable city infrastructure Implementing such technologies is crucial for developing intelligent transportation systems that optimize traffic management and reduce congestion.
This document is for viewing purposes only; any other use requires the prior approval of the National Central Library of Taiwan, the author, and the author's thesis advisor.
Numerous research studies have proposed various techniques to address traffic congestion, often relying on formalized density and velocity characteristics combined within predefined value ranges However, this approach faces limitations because the acceptable range values for density and velocity vary significantly across different roads, leading to potential inaccuracies in congestion assessment and prediction Additionally, current methods tend to analyze traffic data based on fixed rules without real-time verification of actual traffic conditions, increasing the risk of false negative forecasts and unreliable congestion predictions.
This document is for viewing purposes only; any other use requires the prior approval of the National Central Library of Taiwan, the author, and the author's thesis advisor.
Segmented Structure Scheme
Segmented Structure Establishment
Traffic congestion typically occurs in high-traffic-flow road segments when vehicle volume reaches capacity, causing congestion to propagate to neighboring areas Recognizing that each road segment experiences different traffic levels, the authors have developed a segmented structure based on the geographic map of the road network to improve traffic-flow analysis and congestion forecasting Currently, in Taichung City, traffic on main roads is monitored using VD sensors, which provide real-time data essential for effective traffic management.
These roads can be effectively organized as the backbone of a segmented traffic structure To extend this segmented system to other roads, the use of E-tag sensors is proposed, providing a highly efficient solution for vehicle observation E-tag sensors ensure continuous traffic monitoring on non-backbone road segments, enhancing overall traffic management and data accuracy.
The segmented structure is established based on VD and E-tag sensors and
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
The 22 FCU e-Theses & Dissertations (2018), managed by the control center, rely on a comprehensive devices table to facilitate their operations, as detailed in Table 3.1 Segmentation is achieved through the use of VD and E-tag sensors, ensuring precise demarcation for effective system functionality.
3.1.1 VD Sensors Based Segment Demarcation
First, the information of all VD sensors is extracted from Table 3.1 Consequently, latitude and longitude information are used to locate VD devices on a geographical map as shown in Figure 3.1(a)
Figure 3.1 VD Sensors Based Demarcation
The authors divided the geographical map into segments, as illustrated in Figure 3.1(b), to facilitate analysis They assumed that N represents the number of Vehicle Detected (VD) devices on a road, with each device numbered sequentially from 1 to N according to the road's numbering rule This systematic segmentation and device numbering are essential for accurate data collection and analysis in vehicular network studies.
The number of segments on each road is determined by the number of VD devices installed Segment division follows three specific rules to ensure accurate analysis The first rule establishes that the start and end locations of the initial segment are determined using Equations 1 and 2, providing a clear framework for roadway segmentation based on device placement.
This document is designated for viewing purposes only Any other use beyond viewing requires prior approval from the National Central Library of Taiwan, as well as consent from the author and the thesis advisor.
The 23 FCU e-Theses & Dissertations (2018) define the start point of the road using coordinates (𝑥 𝑅 0 , 𝑦 𝑅 0 ), representing its latitude and longitude Additionally, the locations of VD1 and VD2 are specified by their respective coordinates (𝑥1, 𝑦 1 ) and (𝑥2, 𝑦 2 ), which denote their precise geographic positions This coordinate system facilitates accurate mapping and spatial analysis of road networks.
Second, the start location and end location of the last segment can be determined using Equations 3 and 4:
The system calculates the end location (VD_N) using the latitude and longitude coordinates of the last point of the road, denoted as (x_RN, y_RN), which are essential for defining the final segment The coordinates of previous points, (x_N-1, y_N-1) and (x_N, y_N), correspond to VD_N-1 and VD_N respectively, enabling accurate segment identification Normal segments are determined based on rules that utilize the positions of VD_i devices and their neighboring devices along the road Specifically, the start and end locations of a segment where a VD_i device resides are calculated using the latitude and longitude data of the device and its neighboring VD devices, allowing precise delineation of road segments for navigation or analysis.
The distance between two Video Delivery (VD) devices is calculated using their latitude and longitude coordinates, denoted as (x_i, y_i) for the current device VD_i Specifically, the formula considers the coordinates of the current device and its neighboring devices, VD_{i-1} with coordinates (x_{i-1}, y_{i-1}) and VD_{i+1} with coordinates (x_{i+1}, y_{i+1}) This approach allows precise measurement of the spatial separation between devices, which is essential for optimizing network performance and coverage Accurate distance computation based on geographic coordinates is vital for effective device deployment and network planning.
3.1.2 E-Tag Sensors Based Segment Demarcation
For the segment establishment based on E-tag sensors, these sensors are intended
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, the author, and the thesis advisor.
This study highlights the strategic placement of 24 FCU e-Theses & Dissertations at key traffic intersections, such as those near department stores, hospitals, stations, and night markets, to effectively monitor traffic flow To gather accurate traffic data, it is assumed that all vehicles are equipped with electronic transponders, known as E-tags, which facilitate real-time communication Additionally, these vehicles may incorporate GPS sensors to precisely determine their locations on digital maps, enabling comprehensive traffic analysis and improved urban transportation management.
To collect traffic data from E-tag sensors, each road is modeled as a continuous segment These segments are identified using the predetermined locations of two consecutive E-tag readers, enabling accurate monitoring of vehicle flow along each segment This approach ensures reliable data collection for traffic analysis and helps optimize road management strategies.
E-tag readers is responsible for traffic monitoring of one lane of a segment The segmentation is described as follows First, the authors assumed that M is the number of set of devices on a road Second, these set of devices were ordered from 1 to M following the numbering rule of the road Thus, the number of segments on each road depends on the number of set of devices Third, a rule is used to determine segment i segments, where i d_s\), vehicles display stop-and-go flow patterns, where each vehicle's speed is affected by the behavior of the vehicle ahead, leading to a synchronized decrease in average speed across lanes As the density on the road segment continues to rise, the flow \(f\) and density \(d\) become inversely related, with the flow decreasing towards zero as the density approaches the congested density \(d_c\) At the congested density \(d_c\), all vehicles come to a complete stop, and this congestion propagates to neighboring road segments, exacerbating traffic delays and gridlock.
This document is intended for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
32 FCU e-Theses & Dissertations (2018) study, traffic flow factor is modelled according to macroscopic model using collected sensors’ data The data cleansing, traffic flow modeling for VD based segment and
E-tag based segment are described in detail as follows
To guarantee accuracy of the traffic modelling process In this sub-section, the author introduces data cleansing process The process is applied to data collected from
VD and E-tag sensors data cleansing aims to identify and correct errors such as missing information and type mismatches, enhancing overall data quality The process involves ordering records chronologically based on date-time and conducting searches to detect incomplete or mismatched data Each identified defect is then assessed for recovery potential according to predefined relative rules outlined in Table 4.1, allowing for correction or discarding of inaccurate information to ensure reliable data integrity.
Else 𝑃𝑜𝑖𝑛𝑡𝑒𝑟 = 𝑁𝑒𝑎𝑟𝑒𝑠𝑡(𝑉𝐷𝐷𝑎𝑡𝑎𝐿𝑖𝑠𝑡, 𝑃𝑜𝑖𝑛𝑡𝑒𝑟) 2.VD Longitude, VD ID
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, as well as authorization from the author and the author's thesis advisor.
This document is designated for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, as well as the consent of the author and the author's academic advisor.
Relative rules establish relationships between two data fields—namely, the source field and its related field—aimed at recovering missing or damaged information in the source data These rules consist of two components: a type constraint, such as ensuring a value is a positive number, and a logical deductive mechanism that leverages the relationship between fields For example, the lane order in E-tag data has a type constraint requiring it to be a positive number, and it can be reconstructed by using the E-tag ID to deduce the lane order This process involves identifying the nearest related elements within the data set that share the same E-tag ID to facilitate accurate data recovery.
E-tag ID with defected element 𝑖 will be selected Second step is about the nearest element’s lane order data field validating If a validated result is returned, the lane order value of defected element will be assigned as the lane order value of the nearest integrant elements In general, a defected record is recoverable if the missing information or incorrect-type fields can be filled up using the relative rules Otherwise,
This document is for viewing purposes only Any other use requires the prior approval of the National Central Library of Taiwan, the author, and the author's thesis advisor.
In 2018, 35 FCU e-Theses & Dissertations highlighted that defected records will be deleted to ensure data accuracy The final step involves reloading all verified records into the database used for traffic modeling To address data inaccuracies, a defined algorithm for cleansing VD and E-tag data is implemented, ensuring reliable and clean data for analysis.
For VD data cleansing process, the following assumptions are defined to explain the operations
(1) 𝑅𝑉𝐷= {rvd 1 , rvd 2,… rvd m } indicates a set of collected data records of sensors
(2) 𝑉𝐷𝐷𝑎𝑡𝑎𝑙𝑖𝑠𝑡 indicates a list of VD data records which have been ordered according to the value of date-time field.
This document is intended for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, as well as the consent of the author and the author's thesis advisor.
(1) (Line 1) First, a data list is initialized to handle collected data from VD sensors
At the beginning, the data list is set to NIL as no element in the list.
The second step involves sequentially examining the date-time data field of each VD record in the list Each date-time value is converted into an integer that reflects its position within the list Consequently, earlier data records have smaller converted integer values, helping to establish a chronological order for efficient data processing and analysis.
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Traffic Congestion Observation Map
To identify high-risk road segments and provide a comprehensive view of traffic transitions, the authors utilized historical vehicle detection (VD) data for traffic volume analysis to create a traffic congestion observation map They evaluated traffic conditions based on the density and average velocity of each road segment, which are key indicators used in many studies to detect congestion levels Typically, congestion is identified when a road exhibits high vehicle density coupled with low velocity However, this approach faces limitations due to variations in density and velocity ranges across different roads, potentially leading to inaccurate assessments and failed predictions of traffic congestion.
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
44 FCU e-Theses & Dissertations (2018) authors propose the traffic congestion coefficient (TCC) concept, which represents the possibility of a traffic congestion condition existing on a large-scale road network The
TCC is proposed as a measure that is directly proportional to density information and inversely proportional to velocity information, providing a comprehensive understanding of target characteristics To overcome limitations associated with range value classification, density and velocity data are calculated using index metrics methods, enhancing the accuracy of the analysis The TCC is computed through specific equations (Equations 11, 12, and 13), which detail the relationship and calculations involved in the process This approach improves target detection and classification by leveraging both density and velocity metrics effectively.
I d denotes the density performance index;
I v denotes the velocity performance index;
D denotes the average density of the road segment;
V denotes the average velocity of the road segment;
V max denotes the maximum limitation velocity of road segment;
D max denotes the maximum recorded density in the road segment;
V avg _lane k denotes the average velocity of lane k in the road segment;
V max _lane k denotes the maximum limitation velocity of lane k in the road segment;
This document is intended solely for viewing purposes Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
45 FCU e-Theses & Dissertations (2018) l denotes the road segment length; m denotes the number of lanes which have the same travel direction
Because VD data records are extremely numerous, the records should be effectively analyzed to derive useful information Thus, the authors divided historical
VD records into seven sets, one for each day of the week, and then divided each set into
The article details a methodology involving 24 subsets, each representing an hour of the day, to analyze historical VD data records Once classified into these time-based segments, the congestion coefficient value for each segment was calculated to assess traffic conditions accurately Key assumptions were established to streamline the calculation process, ensuring efficient and reliable analysis of congestion patterns throughout the day.
(3) W = {Mon, Tue, Wed, Thu, Fri, Sat, Sun} indicates the day of the week of investigation;
(4) TI = {(0-1), (1-2), …,(23-24)} indicates the time interval of investigation;
(5) TD = {td 0 , td 1 } indicates the travel direction;
(6) 𝑅 𝑖𝑗 = {r 1 , r 2,… r m } indicates a set of historical VD records on day of the week i ∈ W and in time slot j ∈ TI;
(7) 𝑅𝐼𝐷= {rid 1 , rid 2 , …, rid n } indicates the set of road segment IDs;
(8) 𝑇𝐶𝐶 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) indicates the TCC of travel direction td of road segment ridon day of the week i ∈ W and in time slot j ∈ T;
(9) 𝐼𝑁𝐷 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) indicates the density performance index of travel direction td of road segment ridon day of the week i ∈ W and in time slot j ∈ T;
(10) 𝐼𝑁𝑉 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) indicates the velocity performance index of travel direction td of road segment ridon day of the week i ∈ W and in time slot j ∈ T;
(11) 𝐷 𝑚𝑎𝑥 𝑟𝑖𝑑(𝑡𝑑) indicates the maximum recorded density of road segment rid;
(12) 𝑁 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) indicates the number of historical TCCs of road segment ridon day of the week i ∈ W and in time slot j ∈ T.
This document is intended for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
The following TCC calculation algorithm should be executed at the end of the day of the week i and time slot j:
6 If r k device_id = 𝑟𝑖𝑑𝑙 and r k travel_direction=𝑡𝑑𝑝 then
12 For each traveling direction 𝑡𝑑𝑝 ∈ 𝑇𝐷 of 𝑟𝑖𝑑𝑙 ∈ 𝑅𝐼𝐷 do
This document is for viewing purposes only; any other use requires the prior approval of the National Central Library of Taiwan, the author, and the author's thesis advisor.
The variables IND ij rid(td), INV ij rid(td), D ij rid(td), and counter rid(td), used for calculating Traffic Condition Coefficient (TCC) for each road segment, are initialized at the start of the process IND ij rid(td) reflects the density performance index, while INV ij rid(td) represents the velocity performance index D ij rid(td) indicates the density value, and counter rid(td) tracks the number of records for each day of the week and time slot Initially, all variable values are set to zero or NULL, establishing a baseline for subsequent data processing.
Each historical VD data record is analyzed to extract vehicle speed and volume information, which is then classified by road segment These insights provide valuable metrics for understanding traffic patterns and are essential for traffic management and optimization.
𝐼𝑁𝑉 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) and 𝐷 𝑖𝑗 𝑟𝑖𝑑(𝑡𝑑) variables of each road segment.
(3) (Lines 12-15) Third, the previous values of TCC, number of TCC calculation and max density are read for TCC calculation
This document is for viewing purposes only; any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
To enhance the accuracy of the density performance index calculation, the maximum density value is determined by comparing it with the estimated density of the current session If the current density exceeds the existing maximum density, the maximum density value is updated accordingly This approach ensures more precise measurement and monitoring of density performance.
The TCC of each road segment in time slot j on day of the week i is determined using equations 11, 12, and 13 These calculations incorporate both current session data and historical session data to ensure accurate and reliable traffic congestion assessment This approach facilitates effective traffic management by leveraging comprehensive temporal traffic information.
(6) (Lines 23-26) Finally, the calculated TCC value and the number of TCC calculation session of each road segment are stored in the database for the next
The calculation results are stored as tables in the control center, each representing a set of historical VD data Analysis of this data reveals that TCC characteristics are directly proportional to traffic congestion levels, with higher TCC values indicating increased congestion To quantify TCC, road segments are categorized into two types: those with TCC values above 1, indicating a high likelihood of traffic congestion, and those with TCC values of 1 or below, suggesting low congestion or near-free-flow conditions Road segments with TCC exactly equal to 1 exhibit a balance between density and velocity, ensuring stable traffic flow.
This document is intended for viewing purposes only Any other use requires prior approval from the National Central Library of Taiwan, the author, and the author's thesis advisor.
Table 4.2 Traffic Congestion Coefficient Table
ID Location Congestion coefficient value
The control center utilizes data from specific tables corresponding to each time slot and day of the week to reconstruct traffic congestion observation maps These maps highlight areas with a high likelihood of traffic congestion, facilitating efficient monitoring and management of traffic conditions.
Figure 4.2 Traffic Congestion Observation Map