Research On Path Planning Algorithm For Mobile Robots Based On Graph Optimization Laser SLAM

As a novel intelligent device,mobile robots have found extensive applications in various domains including life services,industrial production and military.In order to make mobile robots more autonomous and intelligent,this paper takes LIDAR-based autonomous navigation technology for mobile robots as the research object,and conducts an in-depth study on the SLAM algorithm and path planning algorithm in it.The specific research contents of this paper are as follows.1.The motion model of the differential mobile robot and the observation model based on lidar are established.The occupancy grid map is adopted as the map model in this paper,and its construction principle is theoretically deduced.The ranging principle of lidar is introduced,and a thorough analysis is conducted on the characteristics of collected point cloud data and noise.The radius filtering and voxel filtering methods are used to pre-process the outliers and redundant data in the original point cloud data.2.The basic principles and framework of graph optimization laser SLAM are studied.The Iterative Closest Points(ICP)algorithm is used for scan matching,in order to improve the matching accuracy of traditional ICP algorithms,an improved ICP algorithm is proposed,which integrates the wrong point pair rejection strategy and point-linear feature matching.The scan matching simulation results show that the performance of the improved algorithm outperforms traditional algorithms.Additionally,loop closure detection is achieved using Correlative Scan Matching(CSM)algorithm based on multi-resolution search,which effectively mitigates the issue of cumulative errors caused by sensor errors and other factors.At the back end of the SLAM algorithm,the Levenberg-Marquardt iterative algorithm is used to transform the pose graph optimization problem into a linear optimization problem,and the graph optimization laser SLAM system is completed.3.In path planning,sampling-based path planning algorithms represented by Rapidly-expanding Random Tree(RRT)are studied.These algorithms do not require specific modeling of the map environment and obtain path nodes by random sampling of the environment,but they also have disadvantages such as slow convergence speed and unstable path quality.To address the above problems,this paper proposes an improved RRT* algorithm based on heuristic sampling in path expansion region.First,the greedy search strategy of RRT-Connect is used to quickly search for a feasible path in the map;then,the path is expanded to obtain the heuristic sampling region;finally,the node rejection strategy is used to perform heuristic sampling in the expansion region.Through the Matlab simulation platform,various environment maps are constructed for conducting path planning experiments.The experimental results indicate that the improved algorithm exhibits better convergence speed and path quality compared to existing algorithms.4.To verify the effectiveness of the algorithm in this paper,indoor autonomous navigation experiments of mobile robots are conducted in the simulated environment built by Gazebo platform and the real environment,respectively.SLAM experiments are conducted in the real environment using Turtlebot3 waffle pi mobile robot platform in the indoor long corridor environment and indoor laboratory environment,and by comparing the map building effects,it is proved that this paper map optimization SLAM outperforms the existing Gmapping algorithm and Hector algorithm.After that,the path planning experiments are conducted based on the 2D grid map constructed by SLAM,and the experimental results prove that the EP-RRT* algorithm in this paper is better than the traditional RRT* algorithm and can plan high-quality paths quickly and stably.