A NEW ALGORITHM FOR STATE ESTIMATOR OF QUEUE IN URBAN TRAFFIC COMPUTER CONTROL SYSTEM

Abstract

This paper presents some applications of the discrete-time point process theory, time-varying Markov chain and Martingale Difference(MD) sequence in urban traffic computer control system. In this paper we propose an approach to discrete the place of arriving vehicles in fornt of the intersection, depending on their speed distributions. Furthermore, the formation and dispersion of real queues at traffic signal is then modeled by a time-varying Markov chain which is related to the observation data from vehicle detectors in the intersection. Some results in the theory of discrete time point process are then used to derive a new nonlinear minimum error variance algorithm to predicte the state of queues in the model. At same time, we present a method to determine the time-varying unknown parameters of the algorithm. The new algorithm was tested in comparison with the previous ones by simulation. Some results of the test are shown. They indicated that an apparant better performance than those of previous ones is achived. We have set up an optimal control system based on a micro-computer in one main street in Dalian City, China using the approach presented in this paper. And a satisfactory performance was obtained.