Research On Decision-Making,Planning And Control Of Autonomous Vehicle In High Speed Driving Environment

The development and popularization of autonomous vehicles has provided new thought and method to improve road safety,traffic efficiency and build an intelligent transportation system.As a high-tech integrated set of environmental perception,trajectory planning,decision-making,control execution and information interaction,autonomous vehicle has become a hot research field.How to extract effective information to evaluate the driving status and risks,make a reliable decision in line with the current traffic scene becomes the technical basis for automatic driving;in response to decision-making instructions in different driving environments,how to plan a collision free and executable trajectory by planning system is the key link to achieve automatic driving;while the ultimate goal of automatic driving is to establish an accurate controller to ensure the real-time tracking of the planned trajectory.In this paper,the decision-making,planning and control of autonomous vehicles under high speed environment are studied,including four key topics: intelligent behavior decision-making,car-following execution strategy,lane changing trajectory tracking control and dynamic motion planning.A complete and effective implementation plan of highway environment automatic driving system is proposed.The main contents of this paper are as follows:1.Aiming at the complexity and uncertainty of highway traffic environment,an intelligent safety decision-making model based on deep reinforcement learning and risk correction is proposed.This model extracts the real-time driving information of the subject vehicle and surrounding vehicles to generate decision instructions at each sampling point,and a reward function in the reinforcement learning is designed by considering factors such as driving efficiency and obstacle avoidance.At the same time,the self-attention safety mechanism is introduced into the decision-making framework to improve the decision-making security in complex high speed environment.Then,in order to address the lack of security assurance in the decision-making execution process,this paper designs the risk correction module to avoid the execution of dangerous actions by risk assessment and correction.Through the experiments on the Highwayenv simulation platform,the proposed decision-making model has achieved good results in driving safety and efficiency.2.By using the end-to-end strategy learning method,an autonomous car-following strategy based on driving styles is proposed to execute the following instruction generated by the decision-making module.First,a car-following scenario including the leading vehicle leaves the current lane and other vehicle cuts in line is built,and human drivers’ driving database are collected by the 6 degrees of freedom(DOF)simulation platform.These data contribute to extract the characteristics of different driving styles.Then,based on the consideration of desired car-following distance,driving efficiency and comfort,reward functions are designed to reflect different driving styles,which can be served as a training signal to encourage or discourage behaviors in the context of the car-following maneuver.Finally,the experimental results show that the automatic strategy can integrate the driver’s driving style into the execution process of carfollowing behavior,and also performs well in driving comfort,safety and travel efficiency.3.In regard to the research on trajectory planning and control execution in real highway driving scenes,a dynamic lane change trajectory planning(DLTP)model is proposed and the planned trajectory is tracked and controlled combined with the model predictive control(MPC).The proposed DLTP model combines the dynamic lanechanging trajectory planning algorithm,the lane-changing safety monitoring algorithm and the lane-changing starting-point determination algorithm,which can fully consider the comfort and efficiency of the execution process while ensuring the lane-changing safety.Moreover,the MPC algorithm is introduced as the execution module of the autonomous vehicles.Finally,based on Car Sim-Simulink simulation platform,different high speed driving environments are built by using real-world lane-changing scene data.The experimental results show that the proposed DLTP model is feasible and effective for lane-changing maneuver.4.In order to improve the solution efficiency of trajectory planning process,a dynamic motion planner based on optimization strategy is proposed.An optimal control theory is adopted to select an optimal driving path from the finite path set.The appropriate acceleration and velocity of the execution path are also determined to avoid various types of potential collisions.In order to avoid unnecessary motion replanning process,this study puts forward a collision-avoidance monitoring algorithm to reduce the time consumption of the motion planner.Moreover,an online planning framework based on ‘decision-execution’ is explored.Applying this timeline framework can not only help to evaluate the dynamic planner’s online performance,but also reduce the deviation between the online calculation and the actual execution caused by the time consumption.The simulation experiments in Pre Scan-Simulink platform show that the presented dynamic planner can complete the lane-changing maneuver effectively in different high speed scenes.5.A behavior decision-making,trajectory planning and tracking control system for autonomous vehicles is designed and implemented.The performance of the intelligent system is tested in the highway simulation environment.The experimental results show that the integrated system can realize a safe and fast driving of autonomous vehicles in highway traffic environment,which lays a foundation for further improving the autonomous operation ability of intelligent vehicles.