Research And Implementation Of Intelligent Control Technology To Signal Light At The Intersection

With the development of urbanization and the popularity of the private cars inChina, urban traffic congestion problems have become increasingly prominent. How tosolve this problem in areal, effective and feasible way has attracted wide public concern.Intersection is one of the basic elements of the road network, whose signal control effectdetermines the performance of the entire transport network. Although the traditionalmethod of intersection signal control is quite mature, it cannot adapt to the dynamicnature of urban traffic. The intelligent control method has made certain achievements,but the application of it in practice is relatively small in our country, so the methodneeds to be improved.This thesis firstly analyzes the current signal control method of single crossroads,and discusses the deficiency and shortage of the traditional method under dynamicchanging environment. Then an intelligent control method based on neural network isproposed. Taking into account both the number of queued vehicles in the intersectionand the number of vehicles from adjacent intersection which would be arrived duringthe green phase time, the intelligent method determines the probable green extensiontime. The proposed method, which improved the traditional BP algorithm effectively,optimized the neural network weights by PSO algorithm and formed a BP-PSOoptimization algorithm as a neural network learning algorithm.Based on the aforesaid intelligent method, the method of gradually optimizationcan obtain better control effects. As fuzzy control is particularly applicable to situationwhere the controlled objects are difficult to establish a precise model, and neuralnetwork has learning capacity and a good self-adaptability, combined with theadvantages of that, the green extended controller was designed and the fixed phasesequence fuzzy neural control scheme was formed.To further enhance the effect of signal control, the transformation phase sequencecontrol thinking was investigated. A phase selection controller was designed through theknowledge of fuzzy control and neural network. Combined with the green extendedcontroller in the fixed phase sequence control program, a transformation phase sequence fuzzy neural control scheme was formed.In order to verify the effectiveness and practicality of the control methods proposedabove, simulations of the algorithm are executed with support of VISSIM simulationplatform. The traffic model is built on VISSIM platform, which contains an intersectionused for the confirmation of the three ideas. The control algorithms are coding in C++and main function controls the simulation step via COM interface provided by VISSIM.Statistics data are provided after the simulation and control effect could be seen inanimation. Analysis of results has shown that the intelligent control methods gain abetter control effect than traditional fixed time control, the fixed phase sequence controlscheme is superior to the neural network-based intelligent control method and thetransformation phase sequence control scheme is better than the fixed phase sequencecontrol scheme.