|Ph.D Student||Balasha Tamir|
|Subject||Simulation-based Optimization of Actuated Traffic Signal|
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Tomer Toledo|
|Full Thesis text|
The design of traffic signal control has a profound impact on performance of urban traffic systems. Current traffic signal plans involve complex control logic and a large number of parameters that need to be set. However, little attention has been given to optimization and evaluation of these plans. Simulation-based signal optimization has been limited mainly due to the heavy computational burden associated with it.
This thesis presents the overall structure and the various components of a simulation-based system to optimize the parameters of complex actuated traffic signal plans. It focuses on a development of a mesoscopic traffic simulation - MESCOP (Mesoscopic Evaluation of Signal COntrol Plans). MESCOP is detailed enough to represent the characteristics of actuated traffic signal plans, including representation of the intersection layout and the detectors. The stochastic processes of arrival to the intersection and movement within it are also modeled in detail. The model represents passenger cars, transit vehicles and pedestrians. The system framework incorporates connection between MESCOP, traffic signal design and a genetic algorithm as the optimization method.
The Integrated system to optimize traffic signal plans is demonstrated with an application to a signalized intersection in Haifa, Israel. This intersection is controlled by an actuated traffic signal with transit priority and compensation and queue override mechanisms. The results indicate a large potential to improve the intersection performance, with a reduction of 28% in traffic delays compared to the parameter values set in the original design. Computationally, the results show that MESCOP is very efficient compared to microscopic traffic simulation models, which are often used for similar evaluations. It highlights the benefit of mesoscopic models especially for large scale networks and for optimization processes which require high number of simulation replications.
The system enables to analyze the impact of each parameter within the control plan on the intersection performance by conducting a sensitivity analysis. Local and global sensitivity analyses were applied in this study. The results show that several of the signal parameters do not have any noticeable effect on the model output and therefore can be eliminated from the optimization process in order to reduce its dimensionality.
Additional extensions that were examined include incorporating several secondary objectives and optimization under demand uncertainty. The results show their potential to address the specific needs of the design and to handle demand fluctuations and errors in the traffic measurements.
The system developed in this study is suitable for a wide variety of intersection layouts, performance measures and control strategies, and can help improve traffic signal designs.