Research On Autonomous Vehicle Path Planning Based On Improved A* And MPC

With the rapid increase of global vehicle numbers,social problems such as traffic jam and environmental pollution are becoming more and more serious.Autonomous vehicles have become an emerging research focus because of its great potential in reducing traffic accidents,promoting energy conservation,and establishing smart transportation.Path planning is one of the key technologies in the application of automated vehicles,and it is also a significant subject in the field of the automated vehicle research.Global and local path planning problem is studied based on A* and MPC in this paper.The main contents are as follows:(1)A small scale model prototype is built.The motion form,sensor type and development environment of the prototype are consistent with the real vehicle.The sensing system of the prototype is responsible for collecting information about its own state and surrounding environment,and is mainly composed of GNSS/INS integrated navigation system,LiDAR,IMU and ultrasonic radar.The executive control system mainly includes the main controller,the drive,the actuator and the encoder.The software development environment of the platform is Linux,and the secondary operating system is ROS.(2)A global path planning method based on optimized A* algorithm is designed.On the basis of the traditional A* algorithm,the search space dimension is extended,the node expansion method is improved,and the valuation function structure is optimized.Therefore,the defect of the traditional A* algorithm that automated vehicles motion constraints are not considered is solved.At the same time,the path planning accelerator based on Reeds-Shepp curve is introduced to improve the search efficiency of the algorithm.At last,a path smoother based on cubic spline interpolation is designed to ensure the feasibility of the planned path.(3)An obstacle avoidance path planning method based on MPC is proposed.The discretized vehicle kinematics model is used as the prediction equation.The constraints of vehicle kinematics are given in combination with the actual working conditions.At the same time,the method of selecting the reference point is designed and a new objective function has been developed.In order to verify the realization effect of obstacle avoidance path planning method in the whole working time domain,a path tracking controller based on model predictive control is designed.At last,the joint simulation platform with tracker is built in Simulink/Carsim and the validity of the algorithm is verified.(4)Experiments were conducted at Southeast University to verify the feasibility of the prototype and path planning algorithms.In order to visually and accurately observe the state of the prototype in the experiment,a visual interface was designed and the data format protocol was customized.The experiment first verified the ability of the model prototype to track the path under the automatic driving state,and then added the obstacle avoidance experiment.The experimental results show that the model prototype not only avoids all obstacles,but also maintains the trend toward the reference path.This proves that the model prototype built in this paper and the proposed path planning algorithm are effective and can meet the requirements of vehicle automatic driving.