IN
DEGREE PROJECT THE BUILT ENVIRONMENT, SECOND CYCLE, 30 CREDITS
STOCKHOLM SWEDEN 2020,
Evaluation of Adaptive Traffic Signal Control Using Traffic Simulation
A case study in Addis Ababa, Ethiopia.
AREGAY FKADU KEBEDE
KTH ROYAL INSTITUTE OF TECHNOLOGY
SCHOOL OF ARCHITECTURE AND THE BUILT ENVIRONMENT
Evaluation of Adaptive Traffic Signal Control Using Traffic Simulation:
A case study in Addis Ababa, Ethiopia.
By
Aregay Fkadu Kebede
Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of
Master of Science
In the
School of Architecture and the Built Environment Department of Transport and Geoinformation Technology
Division of Transport and Traffic Planning KTH Royal Institute of Technology
Stockholm, Sweden January 2020
Evaluation of Adaptive Traffic Signal Control Using Traffic Simulation:
A case study in Addis Ababa Ethiopia.
Aregay Fkadu Kebede, 2020.c
Typeset in LATEX
Stockholm, Sweden 2020
Abstract
One of the most significant urban transport problems is traffic congestion. All ma- jor cities both in developed and developing countries are facing the problem due to increasing travel demand caused by increasing urbanization and the attendant economic and population growth. Recognizing the growing burden of traffic con- gestion, community leaders and transportation planners in Addis Ababa are still actively promoting large-scale road constructions to alleviate traffic congestion.
Although Intelligent Transportation Systems(ITS) applications seem to have the potential to improve signalization performance, highly congested intersections in Addis Ababa are still controlled by a timed signal and manual operation. More- over, these pre-timed signal controls are functioning sub-optimally as they are not being regularly monitored and updated to cope with varying traffic demands. Even though the benefits are well known theoretically, at the time of writing of this thesis, Adaptive Traffic Signal Controllers (ATSC) haven’t been deployed in Ethiopia and no research has been conducted to demonstrate and quantify their effectiveness.
This master’s research thesis, therefore, intends to fill the identified gap, by un- dertaking a microscopic traffic simulation investigation, to evaluate the benefits of adopting a Traffic-responsive Urban Control (TUC) strategy and optimizing traffic signal timings. For the purpose of this study, an oversaturated three-intersection test corridor located in the heart of Addis Ababa city is modeled in VISSIM using real-world traffic data. After validating the calibrated model, the corridor was eval- uated with the existing pre-timed, TRANSYT optimized pre-timed plan and TUC strategy. Multiple simulation runs were then made for each scenario alternatives and various measures of effectiveness were considered in the evaluation process.
Simulation evaluation has demonstrated an average delay reduction of 24.17% when the existing pre-timed alternative is compared to TRANSYT optimized plan and 35% when compared to the TUC strategy. Overall evaluation results indicate that deploying the TUC strategy and optimizing the aging pre-timed signal plans ex- hibits a significant flow improvement. It is expected that the result of the thesis work will be an input for future comprehensive policy development processes.
KEY WORDS: Traffic Congestion, Addis Ababa, Ethiopia, ITS, Traffic Signal, Timed Signal Operation, Adaptive Traffic Signal Controller, TUC, Microscopic Traf- fic Simulation, Optimization, TRANSYT, Average Delay.
Sammanfattning
Tr¨angsel ¨ar ett av de st¨orsta problemen i stadsmilj¨oer. P˚a grund av en ¨okad trafikefterfr˚agan har stora st¨ader b˚ade i utvecklade l¨ander och utvecklingsl¨ander stora utmaningar fram f¨or sig. Den fr¨amsta orsaken till ¨okning i trafikefterfr˚agan ¨ar urbanisering, v¨axande befolkning och v¨axande ekonomi. V¨axande tr¨angsel ¨ar ett v¨al k¨ant problem ¨aven i Addis Ababa d¨ar samh¨allsledarna och trafikplanerarna fr¨amjar aktivt storskaliga infrastruktur projekten som en l¨osning till tr¨angsel.
ITS till¨ampning har en potential att f¨orb¨attra trafiksignal prestanda men i Ad- dis Ababa ¨ar h¨og belastade korsningar fortvarande tidstyrda och i vissa fall ¨aven manuellt styrda. Dessa korsningar fungerar inte optimalt i dags l¨aget. Det saknas en regelbunden ¨overvakning och regelbunden uppdateringen som g¨or att korsningar inte kan ta hand om varierad trafikefterfr˚agan. ¨Aven om nyttorna av ITS ¨ar v¨alstuderade och v¨alk¨anda har Etiopien inte till¨ampat n˚agra adaptiva trafiksignal strategier n¨ar detta examensarbete skrivs.
Tesen av denna uppsats ¨ar att fylla identifierade luckan p˚a trafiksignal styrning i Etiopien. Detta g¨ors med hj¨alp av en mikrosimulering som ger m¨ojlighet att upp- skatta nyttor av en adaptiv styrning s˚a som Trafficresponsive Urban Control (TUC) och nyttor av en optimerad styrning i trafiksignalerna. Ett h¨ogbelastat system med tre signalkorsningar i Addis Ababa identifierades som ett l¨ampligt testomr˚ade att simulera i VISSIM. Den validerade modellen anv¨andes f¨or att utv¨ardera f¨oljande sce- narier: befintlig signalstyrning med fasta tider, TRANSYT optimerad signalstyrn- ing och TUC optimerad signalstyrning. Utv¨ardering av dessa scenarier baserades p˚a flertal parametrar som beskriver ˚atg¨ardens effektivitet.
Utv¨ardering av simuleringar visade en minskning i f¨ordr¨ojning p˚a 24.17% i TRAN- SYT optimerad signalstyrning j¨amf¨ort med befintlig signalstyrning. TUC opti- merad signalstyrning visade ¨aven en st¨orre minskning, hela 35%. Generellt antyder utv¨arderingen att b˚ade till¨ampning av TUC optimering och optimering av befintlig tidstyrning ger stora f¨orb¨attringar i trafikstr¨ommar. Resultaten fr˚an denna uppsats f¨orv¨antas inverka framtida riktlinjer f¨or trafikplanering i Etiopien.
Nyckelord: Tr¨angsel, Addis Ababa, Ethiopia, ITS, Trafiksignaler, Adaptiva trafiksig- naler, TUC, Mikrosimulering, TRANSYT optimering, F¨ordr¨ojning.
Dedication
To those kind-hearted people who helped me without the expectation of receiving anything in return. I will always remember your good deeds.
Declaration
I hereby declare that the thesis entitled “Evaluation of Adaptive Traffic Sig- nal Control Using Traffic Simulation: A Case Study in Addis Ababa Ethiopia.“ has been carried out in the Department of Transport and Geoinforma- tion Technology, KTH Royal Institute of Technology, Stockholm, Sweden in collab- oration with WSP Sverige and Stockholm Stad, under the guidance of Xiaoliang Ma, Erfan Aria and Triin Reimal. The work is original and has not been submitted in part or full by me for any degree or diploma at any other University.
I further declare that the material obtained from other sources has been duly ac- knowledged in the thesis.
November 2019 Stockholm, Sweden
Acknowledgements
This thesis work would not have been completed without the contributions and generous support of individuals and institutions. First and foremost, I wish to acknowledge, with a deep sense of gratitude, the master’s degree study scholarship received from the Swedish Institute(SI). Next in line, I would like to thank WSP Sverige and Stockholm Stad for providing the opportunity of writing my thesis in collaboration with them.
This thesis has been conducted under the direction and guidance of Xiaoliang Ma, Docent at KTH Royal Institute of Technology. In addition, valuable advice and continuous support were given by my thesis advisors Erfan Aria (ITS Consul- tant at WSP Sverige) and Triin Reimal (Traffic Signal Planner at Stockholm Stad Trafikkontoret). Hence, I would like to thank them for their patience for the time that the research work has taken, and their spontaneous encouragement and kind deeds. I am also thankful for Christina Diakaki and Aleksandar Stevanovic for their guidance and continuous support.
In the early stages of my research, I was helped by Meskerem Haileyesus, Adonay Kebede and Semhal Gebreselassie during the process of traffic data collection and processing. It is thus my privilege to express my sincere appreciation for their volunteer support. My special thanks go to Lisanu Tebikew and Adisu Wagaw without their contribution, the programming aspect of the thesis work could not be fruitful. I am also indebted to Aydagne Z. Woldemariam and Albania Nis- san(Bibbi) whose, mentorship and expertise have been a great help throughout my career.
These acknowledgments would not be complete without expressing my deepest grat- itude to my families and friends residing in Ethiopia, Sweden, Italy and the USA:
Thank You for everything and I am blessed beyond measure to have you all in my life.
Contents
List of Figures xv
List of Tables xvii
1 Introduction 1
1.1 Background . . . 1
1.2 Problem Statement . . . 2
1.3 Research Objectives and Contributions . . . 2
1.4 Outline . . . 3
2 Literature Review 5 2.1 Fundamentals of Traffic Control System . . . 5
2.2 Signal Timing Optimization . . . 7
2.3 Adaptive Traffic Signal Control Systems . . . 8
2.4 Traffic Simulation Modelling . . . 10
3 Adaptive Traffic Control Method 13 3.1 The TUC Signal Control Strategy . . . 14
3.1.1 Split Control . . . 14
3.2 Traffic System Modelling . . . 22
3.2.1 Microscopic Traffic Simulation . . . 22
4 Computational Experiment and Results 27 4.1 Traffic Simulation Modelling . . . 27
4.1.1 Simulation Model Development . . . 27
4.1.2 Model Calibration and Validation . . . 34
4.2 Scenario Analysis . . . 39
4.2.1 Scenario 1: Existing Field Condition . . . 39
4.2.2 Scenario 2: Coordinated and Optimized Fixed-Time Control (TRANSYT) . . . 40
4.2.3 Scenario 3: Traffic-responsive Urban Control (TUC) . . . 42
5 Conclusion and Future work 49 5.1 Future work . . . 50
Bibliography 51
List of Figures
2.1 Signal Timing Parameters . . . 6
3.1 Functional architecture of TUC . . . 15
3.2 Graph depicting the dynamics of urban road network . . . 16
3.3 Traffic flows in a link . . . 16
3.4 Methodology flow chart . . . 23
4.1 Study corridor modelled in VISSIM (Source: open street map) . . . . 28
4.2 St Estifanos Intersection from VISSIM model . . . 29
4.3 Speed profiles . . . 29
4.4 VISSIM Model . . . 30
4.5 Reduced speed Area specification . . . 30
4.6 Vehicle Composition Summary . . . 30
4.7 Static routing between nodes . . . 31
4.8 AM peak-hour volumes . . . 32
4.9 Traffic control system coding procedure . . . 33
4.10 SA results for various driving behavior parameters . . . 35
4.11 Simulation Result before and after calibration . . . 37
4.12 Speed distribution comparison before and after calibration . . . 38
4.13 Model Validation using speed data (sample distribution curve) . . . . 39
4.14 Existing Traffic Signal Timing Plan for the intersections . . . 40
4.15 TRANSYT Base Model . . . 41
4.16 Signalized Junction from TRANSYT’s base model . . . 42
4.17 Sample phase timing diagram . . . 42
4.18 Cycle time optimization output . . . 43
4.19 Optimized Traffic Signal Timing Plan(Scenario 2 final output) . . . . 43
4.20 Schematic diagram of the signalized urban corridor . . . 45
List of Tables
2.1 Terminologies of traffic signal control [19] . . . 6
3.1 Data sources used for development of microsimulation model . . . 24
3.2 Alternative analysis summary . . . 25
4.1 Typical MOEs used for the study . . . 33
4.2 Web TAG Traffic Flow Calibration Criteria . . . 36
4.3 Optimum driving behaviour parameter values after calibration . . . . 37
4.4 Quantitative Validation result . . . 39
4.5 Network Performance Evaluation Results for all Scenario . . . 47
4.6 Simulation result comparison between scenarios . . . 47
Chapter 1 Introduction
1.1 Background
One of the most significant urban transport problems is traffic congestion. All ma- jor cities both in developed and developing countries are facing the problem due to increasing travel demand caused by rapid urbanization and the attendant economic and population growth. Traffic congestion results in delays, increased fuel consump- tion resulting in increased vehicle operating costs (VOC), increased emissions, and consequent monetary losses and environmental impact.
Different strategies – demand and supply side – are available to alleviate the growing traffic congestion problem and secure more efficient traffic flow and improved mobil- ity. In this light, the Addis Ababa City Administration had only been implementing an extensive road infrastructure expansion project for the last two decades. As ur- ban areas mature, however, opportunities for further investments in transportation infrastructure are often limited and growth in traffic often catches up with upgraded infrastructure. What is more, urban transportation corridors increasingly lack the physical space to accommodate more lanes, resulting in the need for expensive ver- tical infrastructures like over/underpasses and interchanges. Therefore, building infrastructure to create more space and expanding the existing for urban vehicular movement, would lead to further congestion as the induced demand tend to catch up with the increased capacity. The final outcome of policy measures focused on capacity-maximizing measures in most developing Asian cities such as Bangkok, Manila, and Jakarta are cases in point [2]. An alternative approach for improving capacity and mitigating congestion is to optimize the efficiency of existing facilities using Intelligent Transport Systems (ITS) applications. This entails the integrated application of advanced communication, information, and control technologies to improve the efficiency of traffic flow and the effectiveness of transport facilities.
Observation of the high-level congestion patterns in Addis Ababa, in light of the physical and operational bottlenecks, reveals that high-density traffic network with inefficient traffic controls often occurred in urban corridors. In this respect, the in- troduction of ITS service components, notably adaptive traffic signal control (ATSC) would play a vital role in improving urban mobility.
Examining the global practices on road traffic signal control reveals that a variety of innovative traffic control practices are being utilized only in a few large urban areas of developed countries. In addition to the conventional fixed-time controlled inter- sections representing a large portion of signals, the gap between the state-of-practice and state-of-the-art traffic signal control is becoming wide. The two major reasons limiting the wide-scale applicability of last generation traffic signal systems are the huge initial investment and maintenance cost and; the fact that the systems being vendor-specific and delivered in a manner of a “black box”(i.e. lack of transparency and losing control over the controllers).
1.2 Problem Statement
Capacity at intersections is lower than along the road sections leading to them.
Therefore, they constitute the major bottlenecks of an urban road network and of- ten experience long queues and delays. According to statistics, approximately 80%
of vehicle delays on urban roads, and approximately 20% of vehicle emissions occur at signalized intersections [44].
Despite the fact that there have been significant advances in signalization technol- ogy, chronically congested intersections in Addis Ababa are still controlled by a timed signal and manual operation. Moreover, these pre-timed signal controls are functioning sub-optimally as they are not being regularly monitored and updated to cope with varying traffic demands.
Since the extent to which urban traffic is congested is highly dependent on the quality of the signal timing, adjusting poorly timed signals and coordinating traffic signals will significantly improve the traffic flow. This will, in turn, lead to big payoffs in time savings, environmental benefits, and safety. In addition to signal timing opti- mization, upgrading the existing system by replacing the conventional with dynamic and adaptive signal controllers is a fundamental requirement for developing city and notably reduces congestion. However, the concept of traffic signal optimization, advanced signal control strategies and performance evaluation of different control schemes, especially adaptive control approaches has not received significant atten- tion from the Ethiopian research community. To that end, investigations need to be conducted, improved algorithms and optimization procedures should be developed to facilitate the field operations through state-of-practice improvements.
1.3 Research Objectives and Contributions
Recognizing the growing burden of traffic congestion and the importance of efficient access and mobility, community leaders and transportation planners are still actively promoting large-scale road constructions to alleviate traffic congestion. Even though the benefits are well known theoretically, at the time of writing this thesis, Adaptive Traffic Signal Controller (ATSC) has not been evaluated and deployed in Ethiopia, and no research has been conducted to demonstrate and quantify their effective- ness. This master’s research thesis, therefore, intends to fill the identified gap, by undertaking a microscopic traffic simulation investigation, to evaluate and demon-
strate the benefits of adopting a Traffic-responsive Urban Control (TUC) strategy and optimizing traffic signal timings. In pursuit of this goal, the proposed research objectives are to:-
• Devise a methodology to assess different traffic control regimes through mi- crosimulation;
• Collect and process traffic data to supplement the existing available data set;
• Develop a step-wise method to formally build simulation models of traffic intersections;
• Develop a well-calibrated and validated traffic signal evaluation test-bed and;
• Quantitatively assesses the performance of optimized traffic-signal control and adaptive system during oversaturated conditions.
In a nutshell, this research thesis made contributions both in the development of a methodology to accomplish the above-stated objectives and in the actual engi- neering analysis of alternatives (i.e. signal timing optimization and adaptive signal control) for the signalized corridor. Despite the lack of scientific research conducted to address road transport-related challenges in Ethiopia, attempting to tackle the current problem by simulating detailed traffic in calibrated microscopic simulation model is a further contribution of the work.
1.4 Outline
The remainder of the thesis is organized as follows
• Chapter 2 contains important background material and provides an extensive literature review
• Chapter 3 discusses the algorithm behind the TUC strategy and how it is implemented in the microsimulation environment
• Chapter 4 present the comprehensive microsimulation investigation that has been carried out to optimize the existing signal timing plan and to implement and assess the TUC strategy and
• Chapter 5 concludes the thesis by providing concluding remarks, pointing out the significant findings and suggesting future research areas
Chapter 2
Literature Review
To provide a foundation for the thesis, a thorough literature study was conducted. In light of this, chapter two introduces the major findings of the study by emphasizing the elements of traffic control systems and reviewing traffic control-related strategies.
The contents of the chapter include traffic signal control elements, traffic signal control methods, optimization methods, and traffic simulation modeling.
2.1 Fundamentals of Traffic Control System
Traffic signal controllers are the most important instruments for traffic management and operations that ensure the safe and smooth movement of traffic. Primarily, the signal controllers assign right-of-way at intersecting streets or highways to facilitate a continual flow of vehicles and/or pedestrians from every roadway direction [16].
These systems evolved from the first traffic light illuminated in 1862 in London to the state-of-the-art advanced traffic control and management systems. Table 2.1 and figure 2.1 identifies and describes the basic terminology used in this chapter.
Despite the many variations in their design, traffic signals can be classified according to operation mode as pre-timed or fixed, actuated and adaptive. These signals may operate on an isolated intersection (i.e. an intersection located outside the influence of other signals) or coordinated intersections.
Pre-timed or Fixed
Under pre-timed control, the intersection is operated using the same sequence of phases and the length of each phase remains the same for every cycle regardless of the presence or absence of vehicles/pedestrians. Fixed-time traffic signals are ideally used for intersections with fairly constant flows. It may, however, not serve well when the demand is far from stable. This type of plan uses historical data to determine and preset the signal timings. These settings are independent of the current traffic situation; however, some fixed-time control plans use multiple signal setting plans based upon the time of day. Time-of-day pre-timed controls are composed of several pre-timed control plans to satisfy the various traffic volumes during different periods of the day [33].
Key terms Description
Approach A roadway meeting at an intersection. There are two kinds of approaches: incoming approaches in which cars enter the inter- section and outgoing approaches in which cars leave the inter- section.
Traffic Move- ment
Movements of vehicles from an incoming approach to an outgoing approach. The movement comprises a through, left turn, and right turn.
Cycle Time The time required for one complete sequence of signal intervals (i.e. sum of green interval, yellow interval, and red interval).
Phase The portion of a signal cycle allocated to any single combination of one or more traffic movements simultaneously receiving the right-of-way during one or more intervals.
Interval A discrete portion of the signal cycle during which the signal indications (pedestrian or vehicle) remain unchanged.
Split The percentage of a cycle length allocated to each of the various phases in a signal cycle.
Offset The time relationship, expressed in seconds or percent of cycle length, determined by the difference between a defined point in the coordinated green and a system reference point.
Phase- Sequence
A phase sequence is a sequence of phases which defines a set of phases and their order of changes.
Table 2.1: Terminologies of traffic signal control [19]
(a) Typical Speed profiles of vehicles on urban streets [42]
(b) An example of four- phase control [21]
Figure 2.1: Signal Timing Parameters
Vehicle Actuated Traffic Signal Control
Under traffic actuated control, the intersection is operated according to the traffic demands and green time can be extended when traffic is detected. Using sensors to detect and user-programmed settings to control, cycle lengths, splits, and offsets will be subjected to change in response to the detected vehicular and/or pedestrian demand [1]. Actuated traffic signal control can be characterized as fully-actuated or semi-actuated, depending on the number of traffic movements that are detected.
Semi-Actuated: Intersections that utilize semi-actuated traffic signals allow ma- jor movements to receive green unless the signal is alerted by the vehicle detector placed on the minor link. Thus, vehicle detectors are only provided for the minor movements and may be placed on the major movement [37].
Fully-actuated Control: These signals, most suitable for isolated intersections, utilize vehicle detectors for all traffic movements and all phases are actuated. They follow pre-programmed rules to determine which phase to operate given current traffic conditions. Under fully actuated traffic signals, the split, cycle and phase sequence changes from one cycle to another. If vehicles are not detected, it allows phases to be skipped, thereby allowing the controller to reallocate the unused green duration to a subsequent phase.
Adaptive Traffic Signal Control
Adaptive traffic signal control is a method that can adjust signal timings of each intersection on the basis of real-time traffic information from traffic detectors. It’s a dynamic, real-time, on-line approach to reducing traffic congestion by continuously measuring changing traffic patterns and demands. Since the 1970s, several well- known adaptive traffic signal controls have been developed around the world. Some of these control systems are briefly introduced in section 2.3.
2.2 Signal Timing Optimization
Traffic signal optimization is recognized as one of the most cost-effective ways to improve urban mobility. The first analytical method that minimizes the total sum of delays for a two-stage traffic signal was formulated in 1958 by Webster [43]. After the development of the signal timing optimization principles by Webster, many researchers have focused on the development and enhancement of signal timing control and optimization practices [34]. Several analytical computer-based programs have been developed to generate better signal timing plans, including TRANSYT- 7F, SYNCHRO, PASSER II etc. This section provides an overview of three signal optimization tools, Synchro, PASSER, and TRANSYT.
TRANSYT-7F
TRANSYT-7F (Traffic Network Study Tool) is a macroscopic traffic signal model originally developed in the United Kingdom by the Transportation and Road Re- search Laboratory and later modified by the University of Florida Transportation Research Center for the Federal Highway Administration (FHWA) [30]. TRANSYT-
7F is one of the most well developed and widely used fixed time control design sys- tems that optimizes signal timing by performing a macroscopic simulation of traffic flow. This software has been designed, ultimately, for two purposes: simulation of traffic using traffic model and optimizing the traffic signal plan by using signal optimizer [29]. TRANSYT-7F uses a combination of hill-climbing and Genetic Algorithm (GA) optimization methods to optimize cycle length, splits, offsets and phasing sequence on arterials and networks [8]. TRANSYT-7F optimization ob- jective function is expressed in a form of Performance Index(PI) comprising various MOE’s such as delay, progression, stops, fuel consumption, queuing, and through- put [6].
PASSER II
PASSER (Progression Analysis and Signal System Evaluation Routine) II-90 was originally developed in 1974 by the Texas Transportation Institute (TTI) [7]. Only best suited for isolated intersections, PASSER II, like TRANSYT-7F is not a simu- lation model. It operates by utilizing a discrete, macroscopic, deterministic analysis and signal timing model. Cycle length and splits are optimized by Webster’s method and MOE’s and other outputs are generated through the implementation of com- bined algorithms [3]. What makes PASSER differ from SYNCHRO or TRANSYST- 7F is the development of the optimization procedures which lie around the band- width efficiency than of delay. The maximization of bandwidth efficiency is carried out by finding the highest value of summing the thru green band divided by twice the cycle length [8].
SYNCHRO
SYNCHRO is a macroscopic traffic signal timing optimization tool developed by Trafficware Inc. It can be used to optimize signal timing parameters for isolated intersections and to generate coordinated traffic signal timing plans for arterial and networks [8]. SYNCHRO is a delay-based optimization tool in which the objec- tive function configured to minimize stops and delays [47]. SYNCHRO is designed to optimize cycle lengths and splits by Webster’s method and, calculates intersec- tion and approach delays by the method used in the Highway Capacity Manual (HCM). Another significant strength of SYNCHRO is its unique visual displays and user-friendly interface which yields time-space and platoon dispersion diagrams for interactive fine-tuning [3].
2.3 Adaptive Traffic Signal Control Systems
The characteristics of the widely used adaptive signal control systems are briefly summarized below.
ACS Lite
ACS Lite was developed by FHWA, in partnership with Siemens, Purdue University, and the University of Arizona. ACS Lite, a reduced-scale version of the FHWA Adaptive Control Software (ACS), operates in real-time, by making incremental adjustments to split and offset parameters as often as every 5 to 10 min [9]. Cycle
time adjustment is not part of the control module as it is dictated by the time- of-the-day baseline timing plan. The splits adjustments, on the other hand, are made based on the measures of phase utilization [27]. After processing the detector volume and occupancy data during the green duration, the amount of time that the traffic passed across the stop line will be gauged. ACS Lite, in turn, estimates and balances the degree of saturation of each phase [18].
BALANCE
The traffic-adaptive network control BALANCE started in two research projects supported by the European Union (Munich Comfort and TABASCO). BALANCE belongs to the generation of the newest German traffic signal control systems and operates by selecting a signal program with the best cycle time for the current traffic situation from a prearranged set of signal programs. The green split and offsets are determined by the optimization model of BALANCE by applying the Genetic Algorithm while the cycle length is fixed [5].
InSync
InSync is an adaptive traffic signal system developed by Rhythm Engineering (Lenexa, Kansas) that uses innovative sensor technology, image processing, and artificial intel- ligence [36]. The system resulted from the proper integration of the aforementioned elements of InSync optimizes local traffic signals and coordinates signals automat- ically according to the real-time traffic demand. The optimization and operation carried out by InSync bounded within user input thresholds (minimum/maximum parameters) [31].
OPAC
The Optimized Policies for Adaptive Control (OPAC) strategy was originally de- veloped at the University of Massachusetts, Lowell. OPAC is a distributed control strategy involving a real-time signal timing optimization algorithm that calculates signal timings to minimize a performance function of total intersection delay and stops. The algorithm uses measured as well as modeled demand to determine phase durations that are constrained only by minimum and maximum green times and, if running in a coordinated mode, by a virtual cycle length and offset that are updated based on real-time data [31].
RHODES
RHODES (Real-Time Hierarchical Optimized Distributed Effective System) is a real-time traffic adaptive signal control strategy that seeks to optimize the real-time performance of a corridor or network of intersections [20]. RHODES yields an “op- timal” control setting through the network by taking inputs from the detectors and predicting the future traffic throughout the network and yield. Instead of reacting to changes in traffic conditions, RHODES uses peer-to-peer communications and predictive algorithms to identify upcoming changes and prepare accordingly. Thus, there are no fixed offsets and cycle time as both vary depending on the current condition and demand [31].
SCATS
The Sydney Coordinated Adaptive Traffic System (SCATS) is a computer-based and two-level hierarchical adaptive traffic signal control system developed in Australia in the early 1980s by the RTA [25]. It operates, in real-time, by adjusting cycle length, split and offset after the end of every signal cycle throughout the system in response to measured changes of current traffic flow (i.e. detectors output) and system capacity. SCATS manages groups of intersections instead of individual inter- sections in isolation. The intersections controlled in a group are called subsystems.
Since SCATS does not perform optimization, it instead acts as a heuristic feedback system to adjust the signal timings based on changes in traffic flows from previous cycles [26].
SCOOT
Split Cycle Offset Optimization Technique (SCOOT) centralized adaptive traffic signal control system developed in the UK in the early 1980s by the Transport Research Laboratory [22]. SCOOT is often called the online version of the TRAN- SYT signal optimization tool as it continuously measures the traffic demand on all roads in a coordinated network and optimizes signal timings. SCOOT, by using an online model and detectors output, estimates numbers of stops and delays within the network for the next cycle. The adjustment of signal timings is performed in three distinct procedures – the split optimizer, the offset optimizer, and the cycle optimizer. The signal timings for the next cycle will be optimal and reduces the estimated performance measures values (delay and stops) [23].
UTOPIA
UTOPIA (Urban Traffic Optimization by Integrated Automation) / SPOT (System for Priority and Optimization of Traffic) is designed and developed in the 1980s by FIAT Research Centre, ITAL TEL and MIZAR Automazione in Turin, Italy [28].
The objective of the system is to improve the efficiency of both public and private transportation by performing a real-time optimization of the signal timings to mini- mize the total socioeconomic cost of the traffic system. Traffic congestion, vehicular emissions and travel times for public transit vehicles and private traffic are expressed as costs. Depending on the size of the network, the UTOPIA system optimizes the network control strategy over the next 30 to 60 minutes time horizon and updates every 5 minutes. Signal timing optimization at an individual intersection level is performed on the time horizon of the next 120 seconds [31].
2.4 Traffic Simulation Modelling
Computer simulation is playing an important role in the analysis and assessment of freeway and urban street systems, due to its capability and flexibility in modeling traffic conditions, control strategies, and driver behavior [45]. Moreover, simulation modeling is an increasingly popular and effective tool for analyzing a wide variety of dynamical problems which are not amenable to study by other means.
By using simulation techniques, transportation specialists can evaluate alternative treatments, test new designs, conduct safety analysis, train personnel and predict [15]. A variety of traffic simulation models have been developed since the 1960s.
These models are classified based on many factors such as continuity, level of detail, process and problem area. However, categorizing traffic simulation models based on levels of detail, Microscopic, Macroscopic and Mesoscopic, with which they represent the system to be studied is quite common. Macroscopic modeling looks at traffic flow from a global perspective, whereas microscopic modeling, as the term suggests, gives attention to the details of traffic flow and the interactions taking place within it [32]. On the other hand, a mesoscopic model generally represents most entities at a high level of detail but describes their activities and interactions at a much lower level of detail than would a microscopic model [15].
Chapter 3
Adaptive Traffic Control Method
Adaptive Traffic Control Systems (ATCSs) is the third generation of urban signal control systems after pre-timed and traditional coordinated signal systems. ATCSs adjust, in real-time, signal timings based on the current traffic conditions, demand, and system capacity [46]. Adaptive traffic signal systems have been operating suc- cessfully in many countries since the early 1970s. The most widely deployed systems are SCOOT, SCATS, OPAC, UTOPIA, RHODES, TUC, BALANCE, MOTION and PRODYN.
Considering that there are several alternative ATCS, for this thesis work, the Traffic- responsive Urban Control (TUC) strategy has been utilized. Most of the aforemen- tioned ATSCs are delivered in a manner of a “black box”, with little additional information about the control algorithm. Besides other valuable operational char- acteristics, this lack of transparency was the primary reason that leads the author to select the TUC strategy.
Some of the notable characteristics of TUC [10] that fits the objectives of the thesis work are listed as follows:
• It allows straightforward network-wide applicability;
• It is computationally efficient as it permits the execution of all necessary cal- culations in real-time and;
• It is flexible regarding the implementation requirements as it allows an easy installation which can be integrated with the existing systems and infrastruc- tures.
To investigate the effectiveness of deploying ATCSs by comparatively evaluating the TUC strategy and a set of pre-timed signal plans, microscopic traffic simulation is used.
3.1 The TUC Signal Control Strategy
TUC [10] was initially developed in the late 90s as part of an integrated traffic control system (IN-TUC) within the TABASCO (TR1054) project to provide a co- ordinated, traffic-responsive control in large-scale urban networks. The aim of the TUC strategy is to provide, at each cycle, traffic-responsive signal settings by con- sidering the overall traffic conditions within an urban network.
Depicted in figure 3.1, TUC comprises four distinct but interconnected control mod- ules which are complemented by a fifth data processing module. These four modules execute real-time control of green times (split), cycle time, offsets, and the provision of public transport priority by utilizing appropriate methodological tools that follow the principles of automatic control theory, dynamic traffic models, feedback control and optimization techniques [13].
It should be noted that depending upon specific user requirements, any combination of the four control parts may be selected for the application. For example, the user can select to perform only split control, or split and cycle control, etc. The adopted TUC strategy module for this thesis work is Split Control.
The upcoming section of the report aims to describe the basic methodology and char- acteristics of the TUC strategy with an emphasis on the split control module. All the information and illustrations described below are obtained from [10] [11] [12] [13].
3.1.1 Split Control
Controlling green splits was the first part of TUC to be developed and it ultimately aims to minimize the risk of oversaturations and spillbacks of link queues. This split control appropriately manipulates the green times, by using the currently observed traffic loads in the network links, without affecting the given cycle times and offsets.
The split settings for the traffic lights of signal-controlled junctions are produced based on the results obtained from the application of a linear multivariable/regulator control law 3.1 at discrete time instants.
g(k) = gN − Lx(k) (3.1)
Where g(k) is green duration vector, gN is the nominal green times, L is the control matrix and x is the state vector for the number of vehicles within links. Based on Store-and-Forward mathematical modelling approach [17] the multivariable regula- tor is derived by formulating the urban traffic control problem as a Linear-Quadratic (LQ) optimal control problem.
Formulation of LQ Optimal Control Problem
Optimal control consists of theory and methods for controlling dynamical systems so that a mathematical criterion is minimized. LQ Optimal Control Problem is defined by the case when the dynamic system is linear, and the performance index is defined by quadratic functions [24].
The LQ optimal control problem formulation involved a mathematical descrip-
Figure 3.1: Functional architecture of TUC
tion/model of the process to be controlled, the definition of the traffic system con- straints and specification of a performance criterion.
Mathematical Model
To describe the network traffic dynamics as a time-invariant linear system, the urban road network is considered and represented as a directed graph (figure 3.2) with links zZ and signal-controlled junction jJ . In addition, development of the dynamic behavior of the system considers the following assumptions and notations:
• The cycle time Cj, offsets and the total lost time Lj of the junction j are fixed;
• The signal control of junction j is based on a fixed number of stages that belong to the set Fj, while vz denotes the set of stages where link z has right of way;
• The saturation flow Sz, of link zZ, and the turning movement rates tw,z, where wIj and zOj, are assumed to be known and constant;
• Oi and Ij are incoming and outgoing links respectively and
• Cj is equal with C for the sake of simplicity
(a) Urban Road Network
(b) Connecting Link in ur- ban road network [43]
Figure 3.2: Graph depicting the dynamics of urban road network
Considering link z connecting two junctions M and N such that zOM and zIN, the dynamics of link z are given by the Conservation equation. A Conservation Law for traffic flow states that the number of vehicles in a given road link is always computable from the inflow and outflow traffic as long as they are known [40].
Figure 3.3: Traffic flows in a link
By considering a road link depicted in figure 3.3, the vehicle-conservation equation in discrete time can be defined as follows:
x(k + 1) = x(k) + q(k) − u(k) (3.2)
Where x(k) denotes the number of vehicles within link z, q(k) the inflow, and u(k) the outflow, over the time period of [kT, (k +1)T ] with discrete time step index k and control interval T . Using the same concept, to describe the dynamics of connecting link z, equation 3.2 can be rewritten as:
xz(k + 1) = xz(k) + T [qz(k)–sz(k) + dz(k) − uz(k)] (3.3) dz and sz are the demand entering the link (e.g. from parking garages) and the exit flow, respectively. And, for the exit flow the following formula holds true:sz(k) = tz,o qz(k), whereby tz,o stands for the exit rates which is considered to be known and fixed.
Substituting the exit flow formula in equation 3.3 yields:
xz(k + 1) = xz(k) + T [(1 − tz,o)qz(k) + dz(k) − uz(k)] (3.4)
Since turning rates towards link z, tw,z are assumed to be known, the traffic flow into link z is expressed as:
qz(k) = X
wIM
tw,zuw(k) (3.5)
The ultimate benefit of using store-and-forward models for road traffic control is to introduce a model simplification that enables the mathematical description of the traffic flow process without the use of discrete variables. The simplification is introduced when modeling the outflow u of an approach. Provided that space is available in the downstream links and that xz is sufficiently high, the outflow uz of link z is approximated as
uz(k) = SzGz(k)
C (3.6)
Where Gz(k) is the sum of green times on link z involving a constant that may take positive or negative value to consider the r.o.w of link z if extra green time is being utilized during the corresponding stage(s). And, Sz stands for saturation flow. Substituting the inflow and outflow formula presented in equation 3.5 and 3.6, into equation 3.3 yielded the following equation which describes the time evolution of its state in terms of the number of vehicles within it.
xz(k + 1) = xz(k) + T
"
(1 − tz,0)X
wIM
tw,zSw P
ivwgM,i(k) + ew
C + dz(k)
− Sz
P
ivzgN,i(k) + ez C
#
(3.7)
While developing the dynamic behavior of the system, it is assumed that network- wide fixed signal plan gN based on fixed (historical) demands dN is available. This fixed signal control plan gN is also assumed to be optimal in that they lead to steady-state network traffic conditions whereby the developed link queues are close to zero.
(1 − tz,0)qNz + dNz − uNz = 0 (3.8) In line with this, the aim of the multivariable regulator presented in equation 3.1 is to modify these nominal values in real-time to respond to the changes of the demands. Empirically, the above assumption can be expressed by subtracting the steady state equation from equation 3.7
xz(k + 1) = xz(k) + T
"
(1 − tz,o) X
wIM
tw,zSw P
ivw∆gM,i(k)
C + ∆dz(k)
− Sz
P
ivz ∆gN,i(k) C
#
(3.9)
Where ∆gj,i = gi,i− gj,Ni and ∆dz = dz− dNz .
Equation 3.9 captures the dynamics of a connecting link z and for a destination link, since the outflow is not measurable/predictable, it is considered as a disturbance Uz.
For an origin links in the network, since it does not have an inflow qz as outflows from upstream links are not being forwarded,the external demand it receives will be considered as a disturbance Dz. Taking into account the above considerations and adopting the same steady-state version assumption, the equation describing the dynamics of an origin (3.10) and destination link(3.11) is presented as follows:
xz(k + 1) = xz(k) + T
"
(1 − tz,o)∆Dz(k) + ∆dz(k) − Sz
P
ivz ∆gN,i(k) C
#
(3.10)
xz(k + 1) = xz(k) + T
"
(1 − tz,o) X
wIM
tw,zSwP
ivw∆gM,i(k)
C + ∆dz(k) − ∆Uz(k)
#
(3.11) Then, applying equation (3.9), (3.10) and (3.11) to an arbitrary network compris- ing several signalized junctions j J ; the following state equation(in vector form) describes the evolution of the system in time.
∆x(k + 1) = ∆x(k) + B∆g(k) + D∆d(k) (3.12) The real-time application of such control laws requires sufficiently accurate distur- bance forecasts that are not easily obtained in the case of urban traffic. In order to obtain a feedback control law without feed-forward terms that will react to the manifest impact of the disturbances to the controlled process, rather than to distur- bance forecasts, the disturbances are eliminated from the state equation. Where A
= the n x n-state matrix; and B = the n x m-input matrix the state equation reads.
x(k + 1) = Ax(k) + B∆g(k) (3.13)
Objective Criterion
The control objective is to utilize the controlled network homogeneously and keep the developed link queues as short as possible to minimize the risk of oversaturation and the spill-back of link queues. The control objective of queue balancing may be translated into the minimization and equalization of the relative numbers of vehicles within the network links xz/xz,max , where xz,max is the storage capacity of link z Z (measured in vehicles).
A quadratic criterion that considers this control objective has the general form:
= = 1 2
∞
X
k=0
kx(k)k2Q+ k∆g(k)k2R
(3.14) Where Q and R are non-negative definite, diagonal weighting matrices with appro- priate dimensions and they determine the trade-off between minimising queues and adjusting the signal plan. The diagonal elements of Q are set equal to the recipro- cal of the road capacity, 1/xz,max, for road i in order to minimize and balance the relative occupancy of the network links. Furthermore, the magnitude of the control reactions can be influenced by the choice of the weighting matrix R = rI,where I is the unit matrix. To this end, the choice of r may be performed via a trial-and-error procedure in order to achieve the most satisfactory control performance.
System Constraints
The linear-quadratic methodology [10] does not allow for direct consideration of constraints. The control constraints, therefore, will be imposed heuristically after calculation of the feedback law. By definition, the constraint, equation 3.13, applies at junction j where g is the effective green time of stage i at junction j. Additionally, g is constrained to lie in between the minimum and maximum permissible green time, gmin and gmax respectively.
X
iFj
gj,i+ Lj = Cj and gj,i,min ≤ gj,i ≤ gj,i,max; ∀jJ (3.15)
Control Law/Regulator Derivation
Based on linear-quadratic (LQ) optimization technique, the linear multivariable feedback regulator (3.1) is obtained by minimizing the cost criterion (3.14) when it is subjected to (3.15). The application of this optimization technique requires a number of problem simplifications, such as model linearization, quadratic criterion, and no constraints.
The first simplification involves modifying the state equation (3.2) of the overall process to have a generalized form with xRn , uRm and dRp the state, control and disturbance vectors, respectively.
x(k + 1) = f x(k),u(k),d(k),k
(3.16) To facilitate application of linear controller design methodologies, linearization around a stationary nominal point is usual and useful in control engineering practice. Con- sidering the state equation, it can be readily verified that the nominal system steady state just described corresponds to a steady-state form of:
xN = f xN, uN, dN
(3.17) Linearization of (3.16) around the above steady state leads to:
∆x(k + 1) = A∆x(k) + B∆u(k) + T∆d(k) (3.18) where ∆x = x(k) − xN; ∆u = u(k) − uN and; ∆ d = d(k) − dN are the linearized state, control, and disturbance vectors, and A = ∂f /∂x |N, B = ∂f /∂u |N, and T = ∂f /∂d |N, are the state, input and disturbance matrices, respectively. Its is also assumed ∆d = 0 as per the assumption illustrated while driving equation (3.13).
One of the functions which is frequently used to solve optimal control problem for dynamic system is the Hamiltonian. In line with this, recalling the objective criterion, the Hamiltonian function for the problem formulated in (3.16) reads
H
∆x(k), ∆u(k), λ(k + 1), k
= 1 2
k∆x(k)k2Q+ k∆u(k)k2R +λT(k + 1)h
A∆x(k) + B∆u(k)i (3.19)
The necessary conditions of optimality for the formulated optimal control problem are the following:(N.B. lambda is bold)
∂H
∂λ(k + 1) = ∆x(k + 1) = A∆x(k) + B∆u(k) (3.20)
∂H
∂x(k + 1) = λ(k) = Q∆x(k) + ATλ(k + 1) (3.21)
λ(K) = D∆x(K) (3.22)
∂H
∂u(k + 1) = 0 = R∆u(k) + BTλ(k + 1) (3.23) Assuming
λ(k) = P(k)∆x(k) (3.24)
Where k = 0, 1, ..., K − 1. and P(k) is a symmetric n ∗ n matrix, equation (3.23) and (3.24) yields
R∆u(k) + BTP(k + 1)∆x(k + 1) = 0 (3.25) Replacing ∆x(k + 1) from (3.20) in (3.25) and solving for ∆u(k)generates
∆u(k) = −h
R + BTP(k + 1)Bi−1
BTP(k + 1)A∆x(k) (3.26) Introducing the m ∗ n gain matrix
L(k) =h
R + BTP(k + 1)Bi−1
BTP(k + 1)A (3.27)
The following is obtained for ∆u(k) from (3.26)
∆u(k) = −L(k)∆x(k) (3.28)
Replacing (3.20) and (3.24) in (3.21), fetches the following equation
P(k)∆x(k) = Q∆x(k) + ATP(k + 1)A∆x(k) + ATP(k + 1)B∆u(k) (3.29) that, considering (3.28), yields
P(k)∆x(k) = h
Q + ATP(k + 1)A − ATP(k + 1)BL(k) i
∆x(k) (3.30) From which the following matrix equation known as the Riccati equation results for P
P(k) = Q + ATP(k + 1)A − ATP(k + 1)BL(k) (3.31) At last, taking the terminal condition (3.22) and (3.24) into account provides
D∆x(K) = P(K)∆x(K) (3.32)
that is satisfied for
P(K) = D (3.33)
In summary, the time-varying solution of the LQ problem of equation (3.16) is given by the linear feedback law
u(k) = uN − L(k)∆x(k) (3.34)
The time-variant gain matrix L(k) and the time-variant Riccati-matrix P(k) may be obtained by backward integration of the interconnected equations (3.27) and (3.31), starting from the terminal condition (3.33). For most practical applications, a time- invariant solution is more convenient.
To obtain such a solution, the following assumptions are made:
• The problem matrices A, B, Q, and R must be time invariant.
• The time horizon is infinite, i.e. K → ∞.
• The system [A, B] is controllable
• The system [A, F] is observable, where F is any matrix such that FTF = Q Under the above assumptions, it can be proven that the backward integration of the Riccati matrix P(k) starting from any terminal condition P(K) ≥ 0, converges towards a unique stationary value P ≥ 0, thus leading to the following time-invariant control law (3.35) where L is the corresponding stationary value obtained from (3.27) for L(k)
u(k) = uN − L∆x(k) (3.35)
The general solution is provided by the feedback law, where the m∗n gain matrix L is calculated from (3.27) and and P(k) = the solution of the matrix-Riccati-difference equation (3.31).However, for this thesis work, calculation of the time-invariant gain matrix L which entirely depends on the matrices A, B, Q, and R and is obtained straightforwardly as follows
L(k) =h
R + BTP(k + 1)Bi−1
BTP(k + 1)A P(k) = Q + ATP(k + 1)A − ATP(k + 1)BL(k)
Assuming L(k) = L and P(k) = P(k + 1) = P, equation (3.27) and (3.31) will be
L = [R + BTPB]−1BTPA (3.36)
P = Q + ATPA − ATPBL (3.37)
Decomposition of equation (3.37)
P = Q + ATPA − ATPBL P = Q + P − PBL 0 = Q + P − PBL − P
−Q = −PBL P = Q (BL)−1
P = qI(BL)−1 (3.38)
Decomposition of equation (3.36)
L = [R + BTPB]−1BTPA L =h
rI + BT(qI(BL)−1)Bi−1
BTPA
L =h
rI + qBTI(BL)−1B]−1BTP L =h
rI + qBT(BL)−1Bi−1
BTqI(BL)−1 h
rI + qBT(BL)−1B i
L = qBT(BL)−1 rIL + qBT(BL)−1BL = qBT(BL)−1
rL + qBT = qBT(BL)−1 h
rL + qBTi
BL = q BT(BL)−1(BL)
rLBLqBTBL = qBT (3.39)
By solving this quadratic matrix equation (3.39), the control gain matrix can be obtained.
3.2 Traffic System Modelling
Microscopic traffic simulation models are considered to be valuable and powerful tools for analyzing traffic operations in a wide range of planning and modeling tasks. These models are routinely used to evaluate various traffic alternatives and predict performance measures of different operational scenarios [46]. Specifically, they are capable of testing and evaluating intelligent transportation systems (ITS), since many ITS technologies are suited to communicate with individual vehicles and microscopic traffic models emulate traffic flows from the dynamics of individual ve- hicles.
A wide variety of traffic simulation software are currently available and the most widely used packages are CORSIM, AIMSUN, MITSIM, TRANSIMS, VISSIM, PARAMICS, SimTraffic, WATSIM, INTEGRATION, and SUMO. For this thesis work, VISSIM was selected as the environment for the simulation modeling and used to evaluate TUC adaptive traffic signal controller. VISSIM [35] is a micro- scopic, time step and a behavior-based simulation model developed to model and analyze urban traffic and public transit operations. In addition to prior experience of using VISSIM and its technical capabilities, what leads the author to select VIS- SIM as an analytical tool was its Component Object Model (COM) programming interface. The COM interface allows users to develop and implement their own ap- plications on the VISSIM network using a computer programming language such as C++, Visual Basic, or Python.
3.2.1 Microscopic Traffic Simulation
The overall process for developing and applying a VISSIM model involved three major tasks:
• Base Model Development;
• Calibration and Validation, and;
• Scenario Analysis.
A flow chart capturing the overall process adopted to achieve the objectives of the thesis is presented in Figure 3.4. The section below gives a brief overview of the executed sets of procedures that were applied to develop and analyze a validated VISSIM model (refer to chapter four for detail description).
Figure 3.4: Methodology flow chart
Base Model Development
The microsimulation methodology adopted to develop a VISSIM base model to rep- resent the average weekday AM peak period (8:30 -9:30) involved: desk study, data collection & preparation, model coding, and error checking. Before embarking on major analytical effort, the author carried out a desk study which involved defining the scope and objective of the traffic model, reviewing relevant literature, formulat- ing a methodological approach and, selecting an appropriate analytical tool. The input data required for VISSIM have been identified, gathered, and prepared for the study area and the AM peak-hour analytical period. The process of data collection
and preparation started off by identifying and obtaining readily available data from public agencies. And then, it was followed by collecting supplemental data to fill the gaps and update the existing data. Finally, the input data were reviewed to check its consistency and accuracy prior to data entry. Table 3.1 presents VISSIM model data inputs, sources, and uses.
In addition to the basic input data depicted in Table 3.1, developing the base model Table 3.1: Data sources used for development of microsimulation model
Data Source Use
Road Geometry AACRA Input
Traffic volumes Field-collected; AARTMA Input and Calibration Signal Timing Field-collected; AARTMA Input
Speed Field-collected Validation
* (AACRA) - Addis Ababa City Administration Road Transport Management Agency * (AARTMA) - Addis Ababa City Road Authority
required information on vehicle and driver characteristics (vehicle length, maximum acceleration rate, driver aggressiveness, etc.). Upon completion of the data collection and preparation task, replicating the modeled roadway infrastructure in VISSIM was undertaken by building up several layers of network elements. The link and node diagram (infrastructural modeling/network coding) set the foundation for the base model. Additional data on vehicles, traffic controls, route decision, travel demand, simulation run, driving behavior, and etc... were encoded on top of the foundation layer. Before proceeding to model calibration and validation, ensuring the represen- tation of the physical characteristics of the facility, examining its completeness and accuracy were undertaken.
Calibration and Validation
The successful utilization of micro-simulation models is mainly dependent on their accuracy and reliability. Accordingly, it is essential to accurately calibrate and val- idate micro-simulation models before their use to derive critical design decisions.
Calibration of micro-simulation models is the process by which model parameters are adjusted to minimize the differences between the model outputs and the observed values of some traffic parameters [14]. There are no universally accepted procedures for conducting model calibration and validation. In this study, the author adopted the widely used generic procedure which involves: Sensitivity Analysis; Global Cal- ibration; Fine Tuning and Model Validation.
Sensitivity Analysis (SA) - Given the complex and iterative nature of model calibration and a large number of calibration parameters provided in VISSIM (192 parameters, see [35]), SA was performed to identify sensitive operational parame- ters for calibration. An initial screening based on extensive literature studies was performed first to narrow down the numbers of parameters. Then, One-At-a-Time (OAT) SA method was adopted to study the variations in the model outputs due to change in one of the input parameters at a time, while the remaining param- eters are fixed at default values. After properly implementing the OAT method,
model parameters have been prioritized in terms of their effects on model outputs and parameters together with their acceptable value range have been selected for calibration purposes.
Global Calibration and Fine-Tuning - After obtaining a range of values for pa- rameters that cause the model to best reproduce observed traffic counts in the field through SA, the AM VISSIM model has been subject to calibration. Calibrating the model, based on aggregated level, started off by performing global calibration, and then it was followed by link-specific fine-tuning. The first is aimed at adjusting the default driving behavior parameters for typical road sections and the second specializes in fine-tuning site-specific driving behavior parameters at critical loca- tions [4]. For global calibration purposes, the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm was used to optimize the model parameter. The standard method used to compare modeled values against observations on a link, therefore, involves the calculation of the Geoff Havers (GEH) statistic.
Model Validation - After the base model was successfully calibrated against the existing traffic count, the base model was taken forward for model validation. To validate that the traffic model reflects real-world conditions, both quantitative and qualitative checks were conducted on the model outputs for the analysis period.
The quantitative validation process involved a comparison of model output against independent data to ensure that the model is correctly described. To further prove the validity of the VISSIM model, qualitative proof has been carried out to observe the overall traffic patterns of the model such as lane changing behavior, queues, and bottleneck locations and etc.
Scenario Analysis
After the VISSIM model was successfully calibrated and good levels of validation have been reached, it was considered to be an appropriate tool for undertaking the analysis of different alternatives/scenarios. The alternative analysis process
Table 3.2: Alternative analysis summary
Alternatives Description
(1) Existing fixed-time plan
The network configuration, signal timing plan, and traf- fic operational characteristics remained unchanged from the developed base model.
(2) Optimized fixed- time plan
The signal timing settings on the arterial were optimized using TRANSYT
(3) TUC strategy im- plementation
TUC was treated as an external signal controller and its control algorithm have been modeled in VISSIM using the COM programming interface and a python script.
included comparing the existing pre-timed signal timing plan to the optimized plan and traffic responsive control strategy. Table 3.2 presents a brief explanation of the tested scenarios. To evaluate and compare the vehicular traffic operations of all alternatives, the following Measures of Effectiveness (MOE’s) were selected for this study: average delay; average number of stops and; total numbers of stops.
Chapter 4
Computational Experiment and Results
This chapter details the procedure of a microsimulation methodology depicted in Figure 3.4. The first step of the process was partially addressed in the introduction and literature review sections of the report. The subsequent section documents the components of the VISSIM model development, the calibration and validation process and provides a summary of analysis results.
4.1 Traffic Simulation Modelling
4.1.1 Simulation Model Development
This section of the chapter describes how the VISSIM AM peak model network has been built and provides a brief overview of study area characteristics.
Study Area, Data Collection and Preparation
The area used for the ATSCs investigation consists of an arterial (known as Ras Mekonnen Avenue) with three pre-timed signalized intersections. The extent of the study area annotated in Figure 4.1 includes approximately 1.7 km of roadway from the west at Legehar to the east down to St. Estifanos. In addition, the study site experiences heavy congestion as the surrounding land use is a commercial area.
Hypothetical or real-life/field data can generally be utilized to analyze traffic opera- tion through simulations. Hypothetical data are usually exploited when the purpose of the analysis is to conduct a general comparison between traffic operation alterna- tives. For this research thesis, to ensure the simulated site reflect real-world condi- tions, field collected data was adopted. Microscopic traffic simulation process under mixed traffic environment entails the following input data requirements: Supply;
Demand and Control Input. Supply, demand, and control data were obtained from AACRA and AARTMA for St.Joseph and St. Estifanos intersections. However, the data obtained from governmental agencies were not complete enough, consequently, additional data were required to supplement the existing dataset. In line with this,
