A Signaling Game-based Lane-changing Decision-making Mechanism For Autonomous Vehicles

It is well-known that the emergence of autonomous vehicles can not only save road resources and improve travel efficiency,but also ensure road safety in the Internet of Vehicles environment.Restricted by technology,there will be a situation where autonomous vehicles and traditional vehicles are mixed on the road for a long time.Therefore,it is necessary to consider microscopic movement of vehicles under this new type of mixed traffic flow,especially during lane-changing process.However,existing lane-changing models lack analysis of drivers’ heterogeneity,which makes it difficult to provide reference of decision-making for autonomous vehicles during lane-changing process in mixed traffic flow.In order to model lane-changing decision in mixed traffic flow,firstly,this article determines research scenarios and objects,analyzes lane-changing process by introducing signaling game,and sorts out game sequence during lane-changing.In consideration of the heterogeneity of human drivers,the payoff function of each game participant under different lane-changing scenarios is established.In addition,this article solves all the equilibrium solutions and relative conditions of the model based on the refined Bayesian equilibrium theory.Secondly,this paper establishes data cleaning rules according to the characteristics of NGSIM trajectory data.Then,lane-changing decisions are extracted based on phases of lane-changing process and steady state theory.Thirdly,a two-layer parameter calibration framework is established by combining the prediction errors and model equilibrium solutions.Also,this thesis applies Genetic Algorithm and classifier evaluation indicators to iteratively optimize parameters of the model and verify the results of parameter calibration,respectively.Finally,this paper builds a vehicle simulation platform based on Python,which includes road and vehicle initialization,road information update,vehicle micro-behavior control,and model evaluation.Additionally,the MOBIL model is chosen as a comparison to validate performance of the proposed model.The experimental results show that,compared with the MOBIL lane-changing decision model,the signaling game-based model has a better performance in successful rate of lanechanging as traffic density varying.In terms of road safety,the average reciprocals of collision time of the two models are relatively stable and close to zero,which indicates a safe traffic state.In addition,this paper analyzes parameter sensitivity of the proposed model.The results reveal that the signaling game-based model has a relatively good and stable performance under different proportions of aggressive drivers.