Abstract
In traditional coordinated traffic signal control on an urban road network, the following two problems have been known. First, the conventional method involves a time lag between traffic measurement and signal control. Second, an abrupt switching of control parameters throws the traffic flow into disorder.
This paper proposes a new approach to avoid these problems. We increase the frequency of switching. The control parameters are switched as frequent as every two cycles. At the time of switching, minor variations of the on-going plan are generated. For each variation of plan, traffic delay at each intersection is predicted based on measured traffic data at up-stream detectors. Then the plan minimizing the delay is chosen to be the control parameters in the next cycle. In order to evaluate the validity of this approach, experiments were carried out using a traffic simulator. The experiments indicate that the proposed method reduces the queue length significantly, compared with the conventional method.
