Research On Traffic Lights Independent Intelligent Decision Based On Dynamic Bayesian Network

Intelligent traffic light is an important part of the intelligent transportation system.However,the existing traffic lights were powered by timing control,which did not realize the intelligent.In this paper,the artificial intelligence method,which is Bayesian Network theory,is used to establish traffic lights intelligent decision independently model based on Dynamic Bayesian Network.According to the dynamic traffic,a new approximate inference of Dynamic Bayesian Network is proposed to achieve online inference,deciding the optimal traffic lights time.Through the analysis of a large number of data,the main factors that influence the traffic lights time are obtained: whether the trunk road,vehicle speed,vehicle flow,and the correlation with the previous intersection.After discretizing the continuous observed data,the K2 algorithm,which is a structure learning algorithm of Bayesian Network,is used to establish traffic lights intelligent decision independently model with star structure.Then the maximum likelihood estimation algorithm,which is a parameter learning algorithm of Bayesian Network,is utimated to learn parameters,quantitatively depicting strength of the relationship between the variables in the model.In this paper,an approximate online inference of Dynamic Bayesian Network is presented,which is called forward-backward algorithm based on sliding window.The algorithm conbined the sliding window and an exact inference of Dynamic Bayesian Network called improved forward-backward algorithm,through adjusting the window width in the algorithm,the evidence information can be used maximizely to realize online inference.It is proved that the inferring efficiency of the presented inference algorithm is higher after comparing the existing interface algorithm based on sliding window.According to the collected dynamic traffic information,forward-backward algorithm is used to online inference of the model,deciding the optimal traffic lights time in time.The inferred traffic lights time transform correspondingly with the change of traffic information,thus the validity of the model is certifyed.The research result in this paper has important practical significance to solve the problem of traffic congestion and reduce the waiting time of people at the junction with traffic lights.