Research On Driver's Vehicle Route Selection Process Based On The Reinforcement Learning

In recent years,intelligent traffic has been widely used to solve the traffic congestion problem,but whether the published information is valid or producing traffic shocks,depending on the driver's response to the induction of information.When the driver in the face of induction information for path selection,he can choose to accept the induction information,and can also choose to refuse to accept the induction information,which is a game of the process.And because the driver of the path to select the most influential factor is the last strategy to select the proceeds,the income is better,the strategy is strengthened,otherwise weakened,which is consistent with the idea of strengthening the theory of learning.The driver will have the expected return,The probability of the strategy depends on the difference between the expected return and the actual gain.In addition,the reality of the driver of the other drivers of the strategy to choose the belief can not be achieved.Therefore,on the basis of game theory,this paper applies the thought of reinforcement learning theory,to update the probability of the difference between the expected return of the driver and the actual gain.The study of the path selection process of the driver in the face of the induction information is intended to provide the basis for the issuance of the induction information for the traffic manager.Firstly this paper studies the driver path strategy selection process based on the reinforcement learning theory.The traffic induction model based on the reinforcement learning theory is established,the game process of the model is discussed in detail.And the effectiveness of the induced information release is verified by the driver routing algorithm based on the condition of the induced learning information and under the condition of the non-induced information.When the road network traffic changes,the manager should increase the credibility of the induced information to 50%to 60%,Then,based on the established traffic guidance model of reinforcement learning theory,the influence of road network environment change on driver path strategy selection is further studied.Mainly by changing the road network traffic,the initial acceptance of the two factors to analyzes its influence on the driver path selection process.It can make the information is fully utilized,and will not lead to congestion drift,can effectively alleviate traffic congestion;When the initial acceptance ratio is changed,the induction information is issued when the road network traffic accounts for 85%of the capacity and above,that is,when the road network began to congestion,the induction of information can quickly play a role in alleviating traffic congestion on the road to improve road capacity.Finally,the influence of self-parameter change of traffic guidance model on the path selection process of driver is studied.This paper mainly studies the influence on the driver routing process when the two parameters are expected to pay the adjustment speed and the initial acceptance probability.When the pay adjustment speed is expected to change,the expected value of the pay adjustment speed is smaller,the model is more effective,the induction effect is more obvious,the road capacity can be improved rapidly;When the initial acceptance ratio changes,the induction effect of the model is equal to the preference of a strategy when the initial acceptance probability of each strategy is equal,that is,the driver has a preference for a strategy,the model used better,but also meet the real life of the driver is not a layer of the same,but will be divided into various types.