Resarch On Dynamic Obstacle Avoidance Of Intelligent Vehicles Based On Positioning And Environment Mapping

With the research progress of basic theories such as artificial intelligence and machine vision,as well as the upgrading of hardware devices such as sensors and processors,path planning and obstacle avoidance research has always been a hot topic in the emerging blue ocean field of unmanned driving.Considering the complexity of the dynamic environment,this paper studies path planning,positioning and environment mapping and dynamic obstacle avoidance,and proposes the following solutions:On the basis of constructing a grid map,research and optimize the deep reinforcement learning algorithm to realize the global path Planning,improving Kalman filter to improve the accuracy of motion state prediction,verifying the impact of obstacle movement on travel based on cost map,improving artificial potential field method,realizing dynamic obstacle avoidance through local path planning,and finally verifying its effectiveness through simulation platform.First,through the deep reinforcement learning algorithm,the global path planning of the intelligent vehicle is realized.The automatic navigation effect is verified based on Q-Learning and DQN algorithm.This paper proposes to improve the algorithm by optimizing the reward function and exploration strategy,and improve the smoothness of the planned path.Comparing the simulation results,the data shows that the optimized algorithm has faster iterative convergence speed and higher stability,which improves the security and practicability of the planned path.Secondly,on the basis of building a grid map,export a layered cost map,run SLAM to enable positioning and environment mapping functions,and implement a dynamic obstacle avoidance scheme for smart cars based on this.For moving obstacles,call local path planning,determine the risk factor according to the obstacle’s traveling posture,divide the traveling dangerous area,and realize the static representation of dynamic obstacles,so as to realize the dynamic obstacle avoidance by rear detour.Combining the analysis results of the two filters,this paper proposes an extended Kalman filter to improve the accuracy of motion prediction,and optimizes the artificial potential field method to solve the local minimum point problem.Finally,through Gazebo software,simulation experiments are carried out on intelligent vehicle path planning and dynamic obstacle avoidance.In the complex environment built,compare the algorithm operation effect and iterative convergence speed in the global path planning;analyze the optimized local path planning,the success rate of dynamic obstacle avoidance and the time taken.The data shows that the optimized travel route takes into account both safety and efficiency,which proves the effectiveness of the optimization.