Research On Motion Planning Methods For Autonomous Vehicles With Complex Constraints

In order to realize the vision of a safe,efficient,green and civilized smart car power,solve the problems of traffic safety and congestion,and fully meet the people's growing needs for a better life,the task of building smart transportation is increasingly urgent.In recent years,self-driving cars equipped with advanced sensors and other devices and using new technologies such as artificial intelligence have become the focus of attention.This paper takes autonomous driving vehicles as the research object and studies the motion planning problem under complex constraints.This article focuses on two aspects: environment perception and post-decision motion planning.First,study the perception detection layer.Based on the research of the intelligent vehicle obstacle recognition method based on the fusion of lidar and camera,the candidate regions of obstacles are extracted by fusing lidar and camera data.An obstacle recognition model based on convolutional neural network(CNN)and support vector machines(SVM)is proposed.After training data,obstacle targets in visual images are detected.After experimental verification,under the condition of small sample data set,CNN + SVM classifier has better classification results and has stronger robustness.Secondly,in view of the problems in path planning and path tracking in the field of automatic driving,this paper is based on the bidirectional extended balanced joint double-tree RRT algorithm.In order to avoid the advantages of spatial modeling,the objective preference function and the metric function are introduced in combination with the environmental constraints and the constraints of the vehicle itself.At the same time,the regression detection and collision detection mechanisms are combined to solve the local minimum problems in motion planning.Spline interpolation curves generate smooth and continuous executable trajectories,greatly improving the effectiveness of path planning.Next,on the basis of the classic MPC controller,an improved MPC controller based on fuzzy control is proposed.The controller can adjust the weight of the cost function according to the lateral position error and heading error,so that the vehicle can continue to track the path,and at the same time solve the situation that the steering wheel of the vehicle changes greatly in a short time to ensure tracking accuracy and riding comfort.Finally,build a Matlab/Simulink-CarSim autonomous vehicle path tracking co-simulation platform.Experiments show that the improved RRT algorithm can effectively improve the effectiveness of path planning compared to the basic RRT algorithm and the bidirectional RRT algorithm;the improved MPC controller guarantees lateral stability and riding comfort compared to the pure tracking method.