Research On Signal Control And Traffic Flow Prediction For Intersections In Intelligent Transportation System

With the development of society and improvement of living standards,people's demand for trip quality has also increased.More and more people begin to have their own cars,resulting in obvious traffic congestion and bringing great challenges to the urban road traffic.The problem of how to make intersection signal control more effective is attracting more and more attentions from scholars.In this paper,we will focus on improving the performance index of intersection signal control to obtain more effective signal timing based on the traffic status.In addition,short-term traffic flow forecasting is researched to make the signal control real-time and active.The main content is as follows:To begin with,the background and significance,the development of some typical transportation systems and its signal control principles are introduced in the first chapter.An overview of the research state on intersection signal control and short-term traffic flow forecasting is summarized.Afterwards,the signal control on intersections with non-uniform traffic flow is researched.Taking the artery intersection as an example,a real-time optimization strategy with average delay as PI is worked out.Due to the signal control in upstream intersection,vehicles will gather and then travel to the downstream intersection as a platoon.And the average delay will vary according to its arrival time in cycle.On the basis of fuzzy modeling,a comprehensive model aiming at computing average delay of vehicle platoon with an arbitrary arrival time in cycle can be obtained by designing membership functions.Through the artery green wave and analysis on distribution of traffic flow,the parameter in the above comprehensive model,namely platoon arrival time,can be determined and thus make the model feasible in practical application.In the process of optimization,the traditional hill-climbing method is improved in its ability of searching the optimal signal timing.Then three common single indexes are chosen to formulate a multi-objective optimization strategy in the research on intersection signal control with uniform traffic flow.Based on the empirical knowledge,a distribution plan is put forward reflecting how much a single index of the three is valued.Then two different performance indexes which avoid multiple solutions are designed subsequently according to the classification of traffic status.The simulation results using genetic algorithm show that the performance indexes are feasible and helpful to improve the efficiency of intersection signal control.At last,the character of real-time of signal control in present situation is analyzed.Due to vehicles' discreteness,the adaptive scheme based on collecting data by detectors is not true real-time.All the above strategies will be invalid when drastic changes occur in traffic flow.Therefore,a short-term traffic flow prediction method with wavelet decomposition and GMDH algorithm is proposed,which combines the great ability of wavelet in data processing and the powerful ability of GMDH in modelling for complex multivariable system.The simulation results verify that the prediction precision is improved compared with the method just using GMDH or other forecasting algorithms.