Real-time traffic signal control for over-saturated networks
MetadataShow full item record
Traffic congestion results from many sources including bottlenecks, incidents, work zones, bad weather, special events, and poor signal timing. Optimal control of traffic signals has proven to be one of the most cost-effective ways of relieving congestion in street networks. Previous researches on traffic signal timing have been focused primarily on under-saturated traffic conditions. With increasing traffic demands on existing road networks, more and more intersections are becoming over-saturated during peak hours. There is a strong need to investigate the characteristics of traffic flow in congested street networks and develop signal control systems that are capable of system-wide online signal timing optimization. This dissertation is aimed to fill this gap by developing a real-time online signal control system for optimal signal timing of both single intersections and signalized networks under over-saturated traffic flow conditions. A dynamic linear programming model is first developed for single over-saturated intersections. The proposed model provides a real-time formulation procedure to optimize signal timings under given dynamic traffic demand conditions. The research is then extended to network wide signal timing design. A new concept of vehicle detection, gridlock detection, is developed to provide information on backed up queues from the upstream intersections. A heuristic algorithm is then developed to dynamically search the progression routes for the best coordination among the signals in the network. The purpose of the route searching algorithm is to find the best coordination routes to disperse the over-saturated traffic demands to the network intersections. Then, at each intersection level, a compromise approach is developed to optimize the green time allocations simultaneously according to two operational criteria: delay and queue length management. The control concepts and algorithms are then summarized in a comprehensive network signal timing model, namely, the Compromise Network Queue Control (CNQC) model for practical application. The CNQC model features a computer aided, decentralized, and cycle-by-cycle based controlling approach for network wide signal control of over-saturated intersections.