Research On Multiple Robot Exploration And Mapping In Unknown Environment Based On Vision

When the robot carries out military reconnaissance,disaster rescue,search and rescue tasks,it faces the unknown environment,and the exploration and mapping is the premise of its task.In order to change the unknown environment into the known environment,it is necessary to study the exploratory mapping method in the unknown environment.At present,single robot is mainly used for exploration mapping,while the research on multi-robot unknown environment exploration mapping is less.Therefore,based on RGB-D camera,this paper studies the exploration coverage and local point cloud fusion of multi-robot system in unknown environment.The main research work is as follows:In order to solve the problem of repeated exploration when multi-robots explore overlay mapping in unknown environment,this paper studies the extraction of task nodes,task node allocation and exploration path optimization of multi-robot cooperative exploration.Firstly,the frontier algorithm is used to explore the boundary,and the K-means clustering algorithm is used to extract the task node as the exploration node of the robot;Then,the auction mechanism with three cost criteria of exploration path,field of view overlapping and "island" is designed to allocate the multi task nodes and improve the efficiency of multi-robot system cooperation;Finally,when the robot moves to the task node,the probability selection function and logistic model are used to improve the traditional ant colony algorithm and optimize its exploration path.Through the experimental simulation,the exploration method designed in this paper can effectively reduce the repeated coverage rate of multi-robot exploration..For the problem of low efficiency of local 3D point cloud fusion of multi-robot,odometer is used to estimate the initial position and pose,combined with the method of "rough matching+precise matching" to fuse the local images collected by multi robot.Firstly,according to the position and pose of the robot when collecting the point cloud,the transformation matrix model of the local point cloud is established;Then,it is used as the initial iteration value of SAC-IA(Sample Consensus Initial Aligment)rough matching;Finally,combining with ICP(Iterative Closest Point)precise matching for local point cloud fusion,the global three-dimensional color point cloud image is constructed,which effectively improves the efficiency of multi-robot local image fusion.The effectiveness and feasibility of the multi-robot coverage algorithm are verified by the software simulation and robot experiment of ROS system.The results show that the coverage of multi-robot exploration map building is about 98%,and the coverage rate of repetition is about 7%.The method of multi-robot grid map model exploration can effectively reduce the coverage of repeated exploration;The proposed fusion method improves the efficiency of local image fusion.