Research On Multi-Robot Simultaneous Localization And Mapping Problem Based On LiDAR

Robot technology is one of the rapidly developing fields in recent years,involving multiple disciplines such as mechanics,electronics,computer science and control.The development of robot technology has not only promoted the progress of manufacturing,aerospace,healthcare,and other industries but also changed people’s lifestyles and social structure.With the rapid development of artificial intelligence technology,robots have transformed from single-function to multi-function and from single intelligence to multiintelligence,and then possess stronger autonomy and flexibility.In the future,with the continuous emergence of various new technologies,robot technology will have even broader prospect for development.Simultaneous Localization and Mapping(SLAM)is one of the foundations of practical applications of mobile robots,which can help robots autonomously perceive and build maps in unknown environments while accurately localizes their own position.With the mature technology of single-robot SLAM,multi-robot SLAM has also received increasing attention.In multi-robot SLAM,each robot has its own sensors and computing capabilities.They can work together at the same time to generate completer and more accurate map.Coordination of position and attitude between multiple robots is necessary to ensure that they can cooperate with each other and avoid collisions or repeated coverage.In this thesis,simulation experiments are used to study and verify the multi-robot SLAM problem.The definition and mathematical model of the SLAM problem are introduced in this thesis,and a detailed analysis of the robot coordinate system model,kinematics model,laser radar model,and odometer model are provided.At the same time,the theoretical foundation of SLAM based on particle filter algorithm is explained.An improved SLAM algorithm is proposed by this thesis,which utilizes odometer as an auxiliary method to eliminate motion distortion in laser radar data.This method can improve the accuracy and efficiency of the algorithm,thus generating more accurate maps and achieving more precise positioning tasks.In order to verify the effectiveness of the proposed method in this thesis,a simulation environment is built on the Gazebo simulator in ROS,and validation experiments are implemented.By comparing the algorithm before and after improvement,we obtained a more accurate mapping effect.The exploration method based on the Rapid Random Tree(RRT)algorithm is studied by this thesis and is being applied to explore the boundary points of unknown and known areas in the environment.The task allocation algorithm based on the market mechanism is used to allocate the boundary points reasonably to each robot ensure that the multi-robot environment exploration proceeded in an orderly manner and reduce the duplicate exploration of the environment.After the multi-robot is assigned the boundary points,which can ensure path planning is required.Dijkstra global planning algorithm and A* global planning algorithm are compared and analyzed on MATLAB,and A* algorithm is ultimately selected to plan the global path.In addition,to ensure that multiple robots do not collide with obstacles in the environment during the movement process,the Dynamic Window Approach(DWA)algorithm is used to achieve real-time obstacle avoidance.To solve the problem of multi-robot map fusion,image stitching method is designed in this thesis.First,the grayscale processing is performed on the local grid maps explored by each robot,and then the overlap region dabetween multiple grid maps is found through feature matching.Then,these local maps are fused into a global map by a transformation matrix to achieve the fusion of map data for multiple robots.In order to verify the effectiveness of this method,a simulation environment is built on the ROS platform for validation experiments,and real-time map fusion is achieved.To verify the feasibility of multi-robot SLAM,a simulation platform for multi-robot SLAM is constructed by this thesis,which integrated modules for exploring the environment,path planning,map fusion,and SLAM.The platform enabled multiple robots to autonomously explore unknown environments,achieve synchronized localization and map building,output real-time robot pose information,and ultimately output a globally fused map.The entire process is visualized in Rviz.Two simulated environments of different scales and complexities are established in the Gazebo simulator of ROS to verify the feasibility of the multi-robot SLAM experimental platform.