Algorithm Of Fusion IMU Lidar SLAM

The technology of intelligent robot has become a hot spot of domestic and foreign public attention and academic research in the context of the rapid development of artificial intelligence in recent years.Simultaneous Localization and Mapping(SLAM)is an important part in the core perception part of robots,and it is also an area that researchers in the world are currently cultivating.Facing engineering problems such as research,development,application and industrialization of intelligent robots,this paper uses Gazebo simulation platform and Turtlebot3 robot platform as carriers to study and implement mapping,positioning and motion planning in unvisited environments using intelligent robots.The main research contents are as follows:First,systematically model the two-wheels differential mobile robots currently used in the market and research,including robot kinematics models,lidar observation models,IMU measurement error model and map models.Thoroughly analyze and compare SLAM algorithm and motion planning algorithm theory.Secondly,the SLAM algorithm theory based on graph optimization model is derived in detail,and the locating,the mapping front-end and the back end of Cartographer,based on graph-based SLAM,are analyzed in principle.Aiming at solving problem of the intense in robot hardware resources,a distributed Cartographer operation is proposed.The robot only needs to collect sensor data and perform simple control during the process,which reduces the cost and resource intense of the robot.To the Cartographer's front-end low and medium frame rate lidar data,there is a distortion problem.An odometry with higher local accuracy is proposed as an auxiliary to eliminate the Lidar motion distortion and improve the accuracy of the mapping.In view of the short-range error prone to Cartographer Loop Closure detection,lazy decision-strategy is involved,which steadily improves the quality of Loop Closure and can reduce the resource strain caused by calculation.The intelligent robot navigation and motion planning algorithms are analyzed and compared,the theory of the search process of the JPS and TEB navigation algorithms is derived in detail,and the feasibility of the navigation algorithm is verified by experiments.Then,in order to effectively verify the above mentioned algorithms,the Gazebo platform was used to verify the three algorithms of the simulation scenario in turn,which using the 1: 1 model of the Solidworks modeling robot and the simulation environment model imitated normal human house.Correct the IMU used in the experiment.The debugging and verification of the algorithm after performing multiple optimizations on the actual complex unstructured laboratory physical environment with occasional movements.Finally,according to the simulation result of Gmapping,the original Cartographer,the optimized Cartographer algorithm,the results of the optimized Cartographer in complex unstructured environments,compare the performance of them and analyze,based on the quality of the mapping,calculation consumption,and algorithm robustness.The absolute errors and relative errors of the mapping results were analyzed with Matlab.In the cartographer algorithm mapping operation,the global planning algorithm JPS and the local planning algorithm TEB are performed to analyze the feasibility of real-time search paths and the possibility of obstacle avoidance.Experimental results show that the optimized Cartographer algorithm is better than the previous algorithm in performance.