A model entitled SOURCAO (signal optimization under rail crossing safety constraints) is discussed. There are two objectives in SOURCAO design: highway rail grade crossing (HRGC) safety consideration and highway traffic delay reduction around the crossings. The first step in SOURCAO is to choose a proper preemption phase sequence to promote HRGC safety. This is done by the inference engine in an intelligent agent. The second step in SOURCAO is to find the optimized signal phase length to reduce highway traffic delay. The optimization process uses an objective function of the network traffic delay. The delay is approximated and represented by a multilayer perceptron neural network. The system inputs are online surveillance detector data and HRGC closure information. The system architecture of SOURCAO is briefly discussed. Various types of surveillance schemes in the proposed delay model are introduced. Emphasis is on validating the various models applied in SOURCAO. First, the proposed delay model is compared with queue delays obtained from the TSIS/CORSIM traffic simulation package. Second, the neural network delay forecast is verified through a cross-validation procedure. Finally, the objectives of SOURCAO are evaluated by TSIS/CORSIM. The average network delay for 20 runs of simulation is reduced over 8% (P = 0.05; t-test), and the unsafe queue time of the highway vehicle on HRGC in a test case is reduced.
Abstract