Research And Application Of Autonomous Driving Decision Planning Method For Typical Scenarios In The Park

Autonomous driving technology has important application value in park logistics,cleaning,and patrolling.The automatic driving decision-making planning method plays a key role in improving the operation efficiency and safety of intelligent vehicles in park scenarios.In the park,due to the large curvature curve scene,the lateral decision-making results are discontinuous and the offset error is large when the intelligent vehicle is cornering.In addition,the complex obstacle scenes in the park,the weak rules and strong randomness of traffic participants,put forward high requirements for the obstacle avoidance ability of intelligent vehicles.Therefore,this thesis proposes a decision planning method based on the two-point preview driver model and the improved A*algorithm.The research focuses on the three main contents of "two-point preview decision,continuous obstacle avoidance planning,real vehicle platform application".The real vehicle experiments in the campus environment show that the method proposed in this thesis can effectively improve the stability and safety of autonomous driving decision planning in typical park scenarios.The main research contents including:1)In the typical curve scene of the park,in view of the large deviation error of the decision-making result caused by the large trajectory curvature,this thesis proposes a decision-making method based on the two-point preview driver model.By combining the real-time vehicle speed and the curvature information of the global reference path,the position of the preview point is dynamically determined,the deep integration of the twopoint preview theory and the pure tracking algorithm is realized,the decision result of the lateral turning angle is output.At the same time,by combining the road curvature change between the preview points and the expected vehicle speed,the longitudinal speed decision result is output to reduce the offset error of the intelligent vehicle when cornering.In the CARLA simulation environment,the effectiveness of the above decision-making method is verified.2)In the typical multi-obstacle scene in the park,the obstacle scene is complex,and the obstacle avoidance ability of the intelligent vehicle is relatively high.Therefore,this thesis proposes an obstacle avoidance planning method based on the improved A*algorithm.By using the curve coordinate system,the influence of the obstacle position on the continuous obstacle avoidance trajectory is fully considered.Realizing continuous safe avoidance for multiple obstacles.Compared with the traditional A* algorithm,the planning method in this thesis fully considers the obstacle information and improves the safety of the continuous obstacle avoidance trajectory.In addition,by combining the dynamic trajectory planning method and the improved A* algorithm,this thesis realizes the effective avoidance of obstacles in the perception blind area.In the simulation experiment scene,the planning method in this thesis achieves good avoidance for multiple obstacles,which proves the rationality of the planning method in this thesis.3)Finally,this thesis relies on the Hongqi E-HS3 autonomous driving experimental platform of the Robotics Research Center of UESTC.Through the decision-making experiment of the autonomous vehicle in the typical curve scene in the campus environment and the obstacle avoidance planning experiment in the typical obstacle scene.To verify the effectiveness and practicability of the decision planning method proposed in this thesis,a system performance evaluation method is proposed by processing and analyzing the experimental results.To sum up,this thesis mainly studies the automatic driving decision planning method for the large curvature curve scene and the typical obstacle scene in the park.To achieve the purpose of smooth and continuous decision-making results,safe and efficient obstacle avoidance process,and apply the method on the real vehicle platform to provide technical support for the final realization of the automatic driving decision-making planning method.