Best Traffic Path Planning Based On Bayesian Networks

Intelligent transportation system can distribute traffic flow reasonably,ease traffic jam and reduce traffic accidents.And optimum path planning is an important part of intelligent transportation system,which has become a hotspot of the current research.The development of the Bayesian networks(BNs)makes it possible to design a real-time dynamic path planning under the influence of multiple factors.The current existing path planning are mostly only consider a single factor,without taking other traffic factors into account to the decision of making path planning,and can't achieve real-time autonomous decision-making.When it comes to the problem that the existing path planning can't consider synthetically the choice of traffic path under the influence of multi-factor.Path distance,cost,road condition and travel time are mainly factors of influencing traffic path planning.The structure-variable discrete dynamic Bayesian networks decision model of best traffic path is constructed according to real-time observation and related expert domain knowledge.Then adopting the Bayesian networks parameter learning algorithm of maximum likelihood estimation algorithm and parameter adaptive generation algorithm to learn parameters,it determines the quantitative relationship between each influence factor and selected path.Finally,observation data and the quantitative relationship between variables are integrated into Bayesian networks inference algorithm to realize online autonomous decision-making.In view of existing structure-variable discrete dynamic Bayesian networks(DDBNs)inference algorithm can't make an online decision,and structure-variable DDBNs online approximate inference algorithm is proposed,namely single hidden variables structurevariable DDBNs inference algorithm.The concept of units and single hidden variables DDBNs are introduced into structure-variable DDBNs inference algorithm which can quickly online inference with evidence information.The algorithm is compared with filter and structure-variable DDBNs inference algorithm to prove its accuracy and efficiency.The dynamic information of road traffic collected in Xi'an area is taken as the background of simulation application,single hidden variables structure-variable DDBNs inference algorithm is applied to optimal traffic path decision-making model to real-time determine optimum traffic path.Traffic path choice will change accordingly with the change of traffic information so as to verify the validity of the model.It can plan the best traffic path according to different preferences of travelers,which effectively improve the utilization rate of traffic roads.