Research On Mergering Decision-making Model For Autonomous Vehicle Under Urban Environment

Intelligent vehicle is a comprehensive intelligent platform integrating environmental perception and cognition,dynamic planning and decision-making,behavior control and execution,etc.At present,tactical intelligent driving decision-making is one of the key technologies to develop intelligent vehicles.Based on the simulation experiment platform and the reinforcement learning algorithm,this paper studies the adaptive decision-making of intelligent vehicles,which not only can ensure traffic safety,improve the automated decisionmaking ability and have better environment adaptability,but also has important theoretical and practical significance to help intelligent vehicles realize the real road driving.Traditional model-based merging decision-making model is prone to lack of adaptability in complex urban environment.In order to solve this problem,the virtual simulation experiment platform of complex city environment is set up in PreScan,the driver is selected to take the virtual simulation experiment,and the driving data is preprocessed by rough sets.And then,the Q-learning algorithm of reinforcement learning is used to study the driving behavior of car-following,lane-changing and merging.The Reword function is set in consideration of the rules of vehicle operation,safety,comfort,time of merging,and etc.The neural network algorithm is used to generalize the value function(Q value)to improve the ability of Q-learning algorithm for solving the continuity problem.Finally,the driver individual characteristic is proposed to be taken into consideration by two-layer Q-learning algorithm for improving the adaptive ability of intelligent vehicle,and the generalization and validity of the intelligent vehicle merging decision-making algorithm are verified.The results show that,the Q-learning algorithm can improve the decision-making adaptive ability of intelligent vehicle in complex urban environment.The results provide driving knowledge and theoretical basis for the decision-making of intelligent vehicle under complex urban environment.