Research On Driving Behavior Of Autonomous Vehicle In The Mixed Vehicular Platoon At Conflict Road Section

With the increase of economy,the number of vehicles raises,which makes traffic congestion,traffic accidents,traffic pollution and other traffic problems strengthen gradually.Intelligent Transportation System is an effective way to solve road traffic problems,and autonomous vehicle is a key part of Intelligent Transportation System.It can improve traffic efficiency,reduce traffic accidents and alleviate traffic pollution.Vehicular platoon is a hotspot in the field of autonomous driving.The controllability of vehicular platoon can improve road capacity,simplify traffic control complexity,then effectively alleviate traffic congestion,improve road traffic safety and fluency.Due to the high dimensional uncertainty of the driving environment and the limited sensing range of the autonomous vehicles’ sensors,it is crucial for autonomous vehicles that completely,accurately and real-time perceive the change of the surrounding environment.This paper studied a driving behavior decision-making method with perception uncertainty for autonomous vehicle in a mixed vehicular platoon which is driving on a road at the conflict area and consists of connected human-driven vehicles and autonomous vehicle.Firstly,Markov decision theory is used to model the decision-making process of autonomous vehicle.The state features of the vehicle are described from safety,effectiveness and smoothness,and constrained from both kinematics and dynamics.The maximum entropy inverse reinforcement learning algorithm is used to learn the weight of reward function in the Markov decision process from the sample trajectories of human-driven vehicles.Autonomous vehicle obtains additional information by communicating with surrounding human-driven vehicles,and update their confidence values on environment state and social information,in which environmental state is the set of states of all vehicles,and social information is the similar action that some of vehicles takes on the specific situation.The confidence values are incorporated into the driving strategy of the autonomous vehicles.Secondly,dynamic programming is used to solve autonomous vehicle decision process model,and the driving strategy of the autonomous vehicles in the mixed vehicular platoon is obtained.Finally,A road section at the conflict area is set up,comparing the change of speed and headway of autonomous vehicles which adopt aggressive strategy,conservative strategy or the proposed strategy under different settings.The results show that the proposed strategy considering perception uncertainty is separately superior to aggressive strategy and conservative strategy in terms of safety and effectiveness when autonomous vehicle is in the mixed vehicular platoon.