Modelling and simulation of vehicle movements at signalised junctions

• Main Author: Chai, Chen
• Other Authors: Wong Yiik Diew
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Civil engineering::Transportation
• Online Access: https://hdl.handle.net/10356/62243
• Institution: Nanyang Technological University

As an urban environment, Singapore has more than 1,400 signalised junctions. Much attention is paid to maintaining a high service level of the signalised junctions to make sure all road users are travelling efficiently and safely. This research study focuses on microscopic modelling and simulation of vehicle movements at signalised junctions. The study is conducted through four stages. First, traffic data are collected and analysed through field observations and automatic data acquisition. An automatic vehicle classification and tracking method is developed through image and video processing techniques. As vehicle moves in multiple directions at signalised junctions, conventional vehicle tracking algorithms are modified by incorporating a projective transformation to each video frame. Traffic flow characteristics, including traffic volume, arrival distribution, headway, vehicle velocity, and acceleration rates are computed for modelling and simulation purposes. Microscopic movement characteristics, such as moving velocity and acceleration rates are found to be affected by many factors, such as current velocity, front gap, signal phases, and distance to stop line. Cellular Automata (CA) models for vehicle movements at signalised junctions are developed. Compared to existing CA models, the models developed in this study are more flexible for simulating complex vehicle movements at various geometric layouts. To ensure computation efficiency, homogenous and heterogeneous vehicle movements are modelled with two CA models with different cell sizes and transition rules. The proposed models are calibrated and validated both in macroscopic (travel time) and microscopic (velocity profile, acceleration rates and vehicle conflicts) levels. Simulation experiments are conducted to estimate traffic performance in various geometric and traffic conditions. It is found that the performance of geometric design is affected by traffic conditions, including traffic volume and vehicle movements. Apart from relying on quantitative models, microscopic simulation based on CA models can help engineers to assess the traffic performance of their design in various traffic conditions and signal timings. While current CA models are mostly applied on capacity assessment, in this study, proxy indicators, such as “Deceleration Occurrence caused by Conflict” (DOC), are computed in each simulation run to estimate occurrences and severity of vehicle conflicts. Simulated vehicle conflicts shows very good corroborative agreement with accident counts. The proposed safety assessment model is compared to an existing safety assessment method, namely the Surrogate Safety Assessment Model (SSAM). The proposed model is applied in several aspects, including risk degree assessment of different conflict types and estimating safety impact of traffic management strategies, such as permissive right-turn and Red-Light Cameras (RLCs). Compared to existing safety assessment methods based on crash occurrences, CA models are able to replicate realistic vehicle-vehicle and vehicle-pedestrian conflicts and provide safety assessment conditions with user-defined characteristics. Moreover, to simulate driver behaviour, such as perception and decision-making procedures, conventional CA models are incorporated with decision-making techniques. A Fuzzy Cellular Automata (FCA) model and a Neural Cellular Automata (NCA) model are developed with embedded fuzzy sets or Artificial Neural Networks (ANNs). The proposed FCA and NCA models are validated and applied to simulate risky driving behaviour and erratic lateral movements of motorcycles. Compared to conventional CA models, the FCA and NCA models allow users to simulate decision-making procedures of each individual vehicle. It is found that traffic performance (in both capacity and safety aspects) is affected by driver’s behaviour, such as lane-changing, right-turn filtering and risky driving. The perception and attitude of drivers are also found to affect overall traffic performance at the signalised junctions. This study extends current applications of CA models and provides valuable tools and findings for transport professionals in designing and managing signalised junctions. With the addition of new technologies, such as video processing and CA modelling, the relationship between vehicle movements and traffic management strategies are better understood, which is of great value both for academic research as well as practical applications.