Research On Autonomous Navigation Of Driverless Vehicle Based On ROS

Driverless Vehicle are the inevitable trend of future car development.Autonomous navigation is the key technology to realize unmanned driving.It mainly includes two research fields: simultaneous localization and mapping(SLAM)and path planning.SLAM refers to the calculation of the vehicle pose and the construction of the environment map based on the sensor information in the unknown environment map.Path planning refers to planning a smooth,safe and executable motion trajectory to achieve vehicle obstacle avoidance under the condition of vehicle kinematics and dynamics constraints based on the constructed environmental map and real-time perception information.Firstly,the main sensor is compared and analyzed according to different SLAM research methods,and single-line laser radar is selected for the SLAM research in this paper.Then,the advantages and disadvantages of the common map model and the applicable scenarios are analyzed.The grid map is selected as the map model of this paper according to the requirements,and the update principle of the grid map is deduced in detail.Finally,the conversion relationship between the global coordinate system,the vehicle coordinate system and the lidar coordinate system is introduced,and the vehicle motion model and the lidar observation model are constructed.At present,according to different principles,the SLAM method can be divided into two research directions,SLAM based on filter principle and SLAM based on graph optimization.In this paper,the principle of SLAM based on EKF and RBPF is inferred,and the advantages and disadvantages of two kinds of filtered SLAM are analyzed and compared.Subsequently,the tracking performance of EKF and PF in nonlinear systems was tested by MATLAB simulation experiments.Then,the EKF-SLAM algorithm was simulated to analyze the state and map uncertainty.The RBPF-based Gmapping-SLAM method improves the proposed distribution and resampling methods,effectively reducing the number of particles and reducing the rate of particle degradation.Finally,the framework of graph optimization SLAM is introduced,and the principle of Cartographer-SLAM based on lidar is analyzed.This paper divides path planning into global path planning and partial path planning.For the global path planning,the principle and flow of Dijkstra algorithm and A* algorithm based on graph search are introduced in detail.Then the search efficiency and global optimality of the two algorithms are simulated by simulation experiments.Finally,the local path planning algorithm based on dynamic window method is deduced,which mainly includes solving the velocity space,using the velocity motion model to estimate the motion trajectory,and obtaining the optimal trajectory through the evaluation function.Finally,based on the Racecar chassis,the unmanned vehicle experimental platform was built to test the vehicle’s autonomous navigation function.Firstly,the SLAM algorithm is tested in the corridor environment and conference room,and the constructed raster map is compared with the real environment to verify the effectiveness of the SLAM algorithm.Then,testing the vehicle’s path planning algorithm,Racecar safely travels to the target point without collision,and verifies that the A* algorithm and the DWA algorithm can obtain a shorter path with a shorter path that meets the real-time requirements.