Research On Laser SLAM Autonomous Navigation Technology Based On Graph Optimization

As one of the key technologies in intelligent mobile robot systems,autonomous navigation technology is widely used in unmanned vehicles,industrial robots,military reconnaissance,and other fields.Mobile robots use their sensors to determine their position in an unknown environment in real-time while building a map of their surroundings(SLAM)and then realize autonomous navigation functions.In this paper,we focus on two basic problems in autonomous navigation technology(simultaneous localization and map building-SLAM problem,and path planning problem),and design an autonomous navigation system based on ROS(Robot Operating System),choosing Li DAR sensors for environment sensing,and targeting the laser SLAM algorithm of Cartographer algorithm and path planning algorithm based on 8-word encircling track is investigated.In this paper,we first design the autonomous navigation framework,establish the four-wheel differential kinematic model,sensor model,and map model,and provide the theoretical basis for the subsequent research of the SLAM algorithm and path planning algorithm.The physical simulation model of the wheeled robot is built using URDF(Unified Robot Description Format),dynamic and static data conversions are performed between coordinate systems,and the model is completed for validation by comparing the experimental results of the actual and simulated environments.The Cartographer algorithm,based on graph optimization,is theoretically derived and analyzed for the localization and mapping problems.The Cartographer algorithm is compared with the Gmapping algorithm based on filtering according to the different optimization methods,and the mapping effect and CPU usage of the two are compared to show that the Cartographer algorithm is better than the Gmapping algorithm.Considering that Cartographer is prone to map building failure caused by loopback errors in scenes with few and inconspicuous environmental features,the optimization scheme of Map-to-Map loopback detection combined with the delayed decision is proposed based on the original algorithm,and the comparative analysis of map building accuracy and positioning accuracy before and after optimization is conducted in the simulation environment,and the results show that the optimized algorithm has improved map building accuracy The results show that the optimized algorithm has improved the mapping accuracy and positioning accuracy.For the path planning problem,the global and local path planning algorithms are considered,and the global path planning algorithms based on sampling and graph search are simulated for the 8-word encircling track,respectively,and due to the rule restriction of the track with fixed radius turning,the sampling-based RRT(Rapidly Expanding Random Tree)algorithm appears to run the wrong circle in the simulation process,while the graph search can meet the track requirements.After considering the length of the planned path and the number of iterations,the path planned by the A*algorithm based on graph search is determined to be optimal.After obtaining the global optimal path,DWA(Dynamic Window Approach)is used to plan the local path and the implementation of obstacle avoidance function,analyze the influence of the weight coefficients of different evaluation functions on the path planning,and finally determine the optimal combination of weight coefficients.Finally,the effectiveness of the autonomous navigation system is verified by building an autonomous navigation experimental platform,using Velodyne 16-line LIDAR for sense,designing a serial communication protocol,and building a motion control chassis.The localization and map building algorithms are verified in real scenarios,and the functionality of the autonomous navigation system is verified in a simulation environment.There are 55 figures,19 tables,71 references.