An Adaptive Approach for Intersection Vehicle Scheduling with Limited State Information

Vehicle scheduling at the intersection is a challenging topic for the intelligent transportation system (ITS), and various efficient methods have been developed based on different methods. Essentially, compared with the traditional traffic-light-based approach, the improvement of the scheduling performance comes from the increasing information on the states of vehicles collected by the controller. In this paper, we focus on specifying the relative significance of the information for the scheduling performance and propose an efficient vehicle scheduling algorithm when the total information is constrained. Specifically, the scheduling problem is modelled as a sequential decision process, and the required information is the arrival times of vehicles. We first propose a method to determine the decision time sequence by analyzing the state of the intersection. Then an adaptive scheduling algorithm is developed at the decision times, where the information demand is different for each decision according to its relative importance defined by the regret of a wrong choice. The proposed algorithm is verified by simulations that competitive performance can be obtained with much less information required.