Path Planning And Navigation Of Mobile Robot Based On RGBD-SLAM

In recent years,with the development of artificial intelligence and the widespread application of robots,SLAM(Simulation Location And Mapping)as the core technology,can allow robots through their own sensors to establish an environment map and locate it during the exercise when the current environment is unknown.Based on the RGBD-SLAM theory,this paper conducts in-depth research,and proposes a flexible and real-time SLAM-based mobile robot path planning and navigation algorithm for the indoor environment with complex scenes and dynamic moving objects.The main work and innovations of this article are as follows:(1)Understand and analyze the traditional RGBD-SLAM,aiming at the problems of low accuracy and large calculation time of the Bo W algorithm in loopback detection,a fast loopback detection algorithm FAST-Bo W is proposed,using Brief and ORB features and tracking prediction Perform loopback detection.Compared with traditional algorithms,FAST-Bo W recognition rate increased by 12%,and the rate increased by 13%.(2)Aiming at the problem that the traditional SLAM maps lack semantic information,which is not conducive to robots to understand the scene and are susceptible to dynamic objects,the maps have ghosting blur and lack of consistency.The ability of environment perception and scene recognition.Through the pyramid pooling improved MASK-RCNN neural network,the key frames are semantically segmented,and the dynamic target is eliminated by using a lookup table method and semantic information on the segmented key frames.The processed key frames are used to construct a semantic map and perform local bundling adjustment at the same time,and finally perform loop detection to construct a scene map.(3)Aiming at the problems that the A* algorithm has too long search time when planning apath in a dynamic scenario,the path has a large turning point,and it is unable to avoid obstacles dynamically,an improved A* algorithm is proposed.The algorithm performs a second search on the current expansion point,streamlining unnecessary key nodes,reducing path operation points,reducing turning points,improving path generation efficiency,and reducing path length.At the same time,an improved A* algorithm with TEB fusion is proposed to make the A* algorithm perform dynamic obstacle avoidance while ensuring the optimal path.Simulations were carried out on grid maps of different sizes,and the simulation results verified that the improved TEB fusion A* algorithm had better route planning,shorter planning time,and dynamic obstacle avoidance.This paper uses network standard data sets for testing in the improved semantic SLAM system,and compares the map-building effects of different data sets to evaluate the effectiveness and real-time performance of the system.An experimental platform based on ROS(Robot Operating System)was built,and actual scene experiments were carried out on the improved A*algorithm fused with TEB.The experiment proved that the proposed improved algorithm can complete navigation tasks in actual dynamic scenarios.And the path search efficiency is increased by 39%,and the path length is reduced by 4.58%.