Intelligent Vehicle Mobility Model Research Based On Reinforcement Learning

The invention of the automobile has changed people’s living condition and mode,accelerated the communication between the region and the region,and promoted the development of the tertiary industry,such as the upstream machinery manufacturing industry and the downstream sales and maintenance.While bringing convenience to people’s daily life,cars are also accompanied by some major problems,such as traffic accidents and traffic jams on the roads.Only in 2015 the traffic accident caused 1 billion 40 million yuan of direct economic losses and the death of 58022 people,the country’s 60 major city,the peak congestion delay index of the 30 city of more than 1.8,has become one of the main factors restricting China’s economic development and endanger people’s lives and property safety.With the development of sensor technology and VANET technology,Europe and the United States is committed to the development of Intelligent Transportation System(ITS)to solve the above problems.The intelligent vehicle is an important part of ITS system.If the invention of the automobile is a leap in human history of transportation,then the intelligent vehicle can be said that since the invention of the automobile 2.5 century another leap in history of transportation,it free the human driver from the heavy task of driving.The first problem to be solved is the vehicle mobility.The intelligent vehicle has the characteristics of short response time and high precision.The traditional mobile model can not describe the moving behavior of smart vehicles very well.It has become an urgent problem to establish mobile model for intelligent vehicle.This paper studies the mobility model based on reinforcement learning,focusing on intelligent vehicle following model and lane changing model,and discusses the relationship between the mobility model and vehicles surrounding the intelligent vehicle,intelligent vehicle’s motion control,the main contributions are:(1)According to the car following behavior of intelligent vehicle,this paper presents an car-following model based on DQN,and verifies the validity of the model through the simulation platform.(2)For lane changing execution behavior of intelligent vehicle,designed the ideal reference trajectory,trajectory tracking based on the control of QSMC,and the Actor-Criticalgorithm combined with QSMC control makes the lane change execution of intelligent vehicle more practical.The simulation results show that the process is stable and comfortable.