With the continuous improvement of artificial intelligence,mobile robots are being used more and more widely,and improving the intelligence of mobile robots has become an important research topic.This thesis focuses on semantic map construction technology based on 2D LIDAR and depth camera fusion,and path planning technology for autonomous navigation of mobile robots,aiming to develop a mobile robot working in indoor scenes with semantic map construction and autonomous navigation function.The main research contents of this paper are as follows:(1)An indoor omnidirectional mobile robot was built as an experimental platform based on the Mc Namee wheel chassis,equipped with sensors such as depth camera,2D Li DAR and rotary encoder,and the forward and inverse kinematic models of the Mc Namee wheel chassis were analyzed.An odometer-based motion model was established for the motion process of the robot system,and a 2D Li DAR-based observation model was established for the observation process.The software platform was built under Linux operating system using the robot operating system to realize the functions of remote control,semantic map construction and autonomous navigation of the robot.(2)A semantic grid fusion map construction method for mobile robots is proposed.A twodimensional occupancy grid map is constructed using the LIDAR-based RBPF-SLAM algorithm as the basic framework of the fusion map.The semantic and coordinate information of objects in key frames is extracted by a deep learning target detection algorithm.The semantic grid fusion map is constructed by adding semantic markers to obstacles using 3D coordinate transformation of depth camera mapped to a 2D occupancy grid map.(3)The RRT~* algorithm for global path planning of indoor mobile robots is improved.In order to solve the problems of strong randomness of RRT~* algorithm expansion,slow convergence speed and many inflection points of generated paths,a double-tree expansion strategy with target bias is introduced to speed up the random tree expansion and find the initial path quickly.The idea of randomized sub-region optimization is used to optimize the planned initial path,and the simulation results show that the improved RRT~* algorithm reduces the number of samples in useless regions,improves the optimization efficiency,and can complete the path planning faster.(4)A two-layer path planning method is proposed by fusing the improved RRT~* algorithm with the dynamic window approach.The improved RRT~* algorithm is used to plan the global optimal path,and the global path nodes are used as local sub-target points to guide the dynamic window approach for local path planning with adaptive control of heading weights to achieve dynamic obstacle avoidance of the robot.In multi-objective point path planning,the improved RRT~* algorithm and ant colony algorithm are combined to solve the multi-objective point path planning problem for mobile robots. |