Research On Lane-Changing Decision Of Autonomous Vehicles In The Connected Environment

As traffic congestion,environmental pollution,traffic accidents and other traffic problems become increasingly prominent,unmanned technology is considered to be one of the effective ways to solve the current traffic problems.The mobility model is an important part of the unmanned technology and is the basis for decision-making and control of the autonomous vehicle.The main tactical decisions and behaviors of the autonomous vehicle are lane-changing decisions and the car-following.The lane-changing decisions are the main source of collisions and congestion,and they are the most challenging tactical decisions.However,many change models are based on the mathematical formula and traffic flow theory.Although they can simulate the lane-changing process of vehicles and provide decision to a certain extent,the interaction between vehicles and other factors are not considered,which make the decision-making performance lower and can not be used as a decision-making for lane-changing of the autonomous vehicle.This thesis mainly studies the lane-changing decisions of the autonomous vehicle in the connected environment.The main contributions are as follows:(1)The lane-changing scene is regarded as a non-cooperative game under complete information.By analyzing the game elements and behaviors in the lane-changing scenario,the lane-changing decision of the vehicle is proposed and simulation experiments are conducted for the lane-changing decision of autonomous vehicles in hybrid scenarios,which joints driving style and game theory.Finally,the effectiveness of the lane-changing decision is verified on the data set.Introducing the style coefficient into the game payoff of the parties reflects the important influence of different styles on the payoff,which is is more coincident with the actual situation.The constant acceleration and the discrete acceleration of the intelligent driver model are used to predict the position of the lane-changing vehicle,which is closer to the actual situation.This decision model can be further used in the development of driver assistance systems and as a lane-changing decision for autonomous vehicles in the mixed environment of manned and autonomous vehicles.(2)A method of lane-changing decision based on function approximation Q-learning is proposed for the autonomous vehicle.Using the headway to reduce the dimension of state space is beneficial to the exploration and learning of strategies,and setting the explorationfactor of the ?-greedy strategy reasonably to speed up the convergence speed of the algorithm.At the same time,using function approximation to realize the generalization of reinforcement learning,which can solve the real problem better.Finally,simulation experiments are carried out and the method of function approximation Q-learning is used to control the lane-changing of the autonomous vehicle.Furthermore,the effectiveness of the method is verified by comparing with above method and other method.