Model Free Adaptive Multi-intersection Coordinated Signal Control Based On RBF Neural Network

In recent years,with the development of urbanization and the continuous improvement of people’s quality of life,private car travel has become the mainstream mode of transportation.Subsequently,there are issues such as traffic congestion,resource waste,and safety.The cost of alleviating these problems through infrastructure construction is too high and the cycle is long.Therefore,on the basis of existing transportation facilities,various algorithms have become the main means to achieve more reasonable control of signal lights.At the same time,current facilities can obtain a large amount of real-time traffic data from roads.If these data are used to alleviate traffic,it can alleviate traffic problems at a lower cost.In addition,the actual traffic flow in daily life will have significant fluctuations on a periodic basis,and control algorithms generally targeting specific traffic flow are no longer applicable.In this context,this article focuses on three types of intersection signal control problems and does the following work:Firstly,aiming at the problem of signal control at multi-phase single intersection,this thesis proposes a queue-balancing algorithm based on model free adaptive control(MFAC)by using full-format dynamic linearization(FFDL),and using radial basis function neural network(RBFNN)to act on parameters of MFAC to make them adaptively adjusted to cope with the rapidly changing traffic flow.Thus,FFDL-RBFNN algorithm is formed.In addition,a variable period algorithm is added to reduce the time loss effectively.The algorithm is applied to four-phase single intersection with time-varying traffic flow.Secondly,in view of the coordination problem of trunk lines,this thesis also introduces the idea of decentralized estimation and decentralized control.Combined with the FFDL-RBFNN algorithm mentioned above,the algorithm can be updated into the decentralized estimation and decentralized control FFDL-RBFNN,and the algorithm is applied to the coordinated control of multiple trunk intersections with green wave belt on the basis of fixed period.Moreover,these intersections are four-phase intersections with time-varying traffic flow,and the two phases of non-trunk lines are fixed timing.Then,aiming at the problem of coordinated control at multiple intersections,this thesis proposes a multi-agent system.Agents are set at each intersection to realize information interaction between different intersections,and the decentralized estimation decentralized control FFDL-RBFNN algorithm is set in multi-intersections to carry out regional multi-intersections coordinated control.At the same time,the decentralized estimation decentralized control FFDL-MFAC is proposed,and the algorithm is also applied to the coordinated control of four-phase multi-intersection with time-varying traffic flow,and the above two proposed algorithms are compared in simulation with other two traditional control strategies,to show the effect of the proposed algorithm.Finally,the proposed algorithm is analyzed with the actual intersections and regions of Daxing District,Beijing as the research object,and compared with the traditional control strategy respectively under the condition of different traffic demands.According to the simulation results of queue length at each phase and the average time consumption of vehicles,it can be seen that the proposed algorithm controls at a single intersection.Coordinated control of trunk intersections and coordinated control of regional multi-intersections can effectively reduce the length of vehicle queuing and the time consumed by vehicle passage,and improve travel efficiency.