The Optimization Research Of Based Td Method In Signal Control Of Single Intersection

In recent years, traffic congestion has become an important factor in restricting thesocial and economic development, and gets more and more attention. Traffic signalcontrol systems have become an important measure and means of managing networktraffic flow. Optimizing it has become an important way to improve the efficiency oftraffic operations.This paper adopts the TD learning algorithm in reinforcement learning to studythe single intersection signal control, which is a useful attempt to improve controlsystem performance. The purpose is to put forward intersection signal control newmethod and new thinking.The saturation concept of the green time extension, then proposes classification ofgreen time saturation. On this basis, for the classification of green time saturation target,construct a reward function, using state of the discrete method and fuzzy method tosolve the traffic state space dimension explosion problem, the establishment of eightfixed period and variable cycle two modes offline TD learning with optimizationmodel. Through the Matlab programming, the development of the eight kinds of modelcalculation procedures, compared with the on-line learning model for off-line learningmodel is more suitable for intersection signal timing optimization. Using a two-phasecontrol of a single intersection as an example, the comparative analysis of theperformance of eight models optimized。General fuzzy TD model outperforms the stateof TD discrete model, variable cycle mode offline TD learning model can be obtained,the structure of optimal solutions distribution, which is the traditional timing theorydoes not have. Periodic conditions, reward grading effect; variable cycle condition,reward grading effect is obvious, the traffic better performance.