Research On SLAM And Path Planning Of Mobile Robots In Complex Environment

In order to implement the key tasks of the " Fourteenth Five-Year Plan " for the development of the robot industry and improve the ability of robots to promote high-quality economic and social development in China,it has become an important goal for future scientific and technological development in China.As an important role in the entire robotics field,mobile robots have autonomous navigation functions that are the foundation for robots to intelligently complete complex tasks.In response to the problems of low algorithm efficiency and high memory consumption in the two important links of SLAM and path planning during the autonomous navigation process of robots,this paper implements a low memory consumption real-time mapping solution based on the RTAB-map algorithm,and improves the traditional A* algorithm to propose a more efficient and memory efficient path planning solution.The main research content is as follows:(1)Establishment of a motion and perception model for a dual wheel differential mobile robot.The kinematics model of the robot is established based on the structure of the differential mobile chassis,and the relationship between the control speed and the motion behavior of the robot is obtained;Based on the principle of lidar measurement,filters are used to denoise radar noise points generated by jitter and occlusion effects in radar data;Based on the principle of camera imaging,calculate the conversion relationship between coordinate systems,calibrate the internal parameters of the RGB-D camera,and correct the distortion of the RGB-D image based on the principle of image distortion;Establish a Gazebo simulation environment based on the ROS system framework,robot structure model,and sensor model,providing conditions for subsequent research on SLAM schemes.(2)Research on joint mapping of Li DAR and RGB-D cameras.Statistical filtering and voxel filtering were performed on the dense point cloud obtained by RGB-D cameras to reduce the number of noise points in the point cloud,and joint calibration and data fusion of lidar and camera were achieved;Based on the RTAB map algorithm,the memory management mechanism is used to realize the real-time construction of dense point cloud map,and the obstacle point clouds on the ground and higher than the robot range are segmented and filtered.The results show that the octree map built based on the optimized point cloud map occupies 1% of the original map memory,and the two-dimensional map obtained from this can more accurately express the obstacle information in the environment,which is more suitable for mobile robots to carry out path planning.(3)Research on improving global and local path planning algorithms.Improve the global algorithm based on the feature of redundant nodes generated during A* algorithm path finding:improve the heuristic function by vector cross multiplication and introduce a balance factor to increase the search trend of the algorithm in the target area;By combining the skip search strategy,the algorithm achieves variable step size skip search,avoiding the drawbacks of low efficiency in layer by layer search.We improved the sampling path evaluation function of the DWA algorithm,set a distance threshold to determine the timing of local target point conversion,and proposed a local target point selection strategy,achieving the fusion of the improved global algorithm and the improved local DWA algorithm.The simulation results show that,under the condition of basically equal path lengths,as the size and complexity of the map increase,the pathfinding efficiency and memory consumption performance of the improved algorithm continue to improve.Compared with A* and JPS algorithm,the search efficiency in 100×100 grid map has increased by 83.75% and 72.48% respectively,and the number of search nodes is reduced by 98.40% and 63.53% respectively.After fusion with DWA algorithm,it has good dynamic obstacle avoidance performance and smoother paths.(4)Implement mobile robot mapping and autonomous navigation.The mobile robot hardware platform and software framework are constructed to verify the mapping and navigation effect of the algorithm in the actual environment.The results show that: the octree map obtained by the SLAM scheme proposed in this paper accurately represents the three-dimensional space occupation state,with less memory occupation,and the two-dimensional grid map obtained from this is more accurate in describing the environmental information during path planning;Compared with traditional A* and JPS algorithms,the improved global path planning algorithm has significantly improved path finding efficiency and memory consumption,and the improvement effect is directly proportional to the map size and complexity.The simulation results have been verified,and the good fusion effect of the two improvement strategies has also been verified from the perspective of path quality.After combining the proposed fusion strategy with the improved DWA algorithm,it has good overall adaptability to dynamic environments.