Research On Development And Application Of Graph-SLAM System For Mobile Robot

With the continuous advancement of technology,robotics technology also has been greatly developed.As an important member of the robot family,mobile robots are being widely used in industry and service industries,but they still face some challenges in autonomous mobility.In recent years,the ability about perceiving the surrounding environment of mobile robots has been given with the development of the sensor-oriented SLAM technology,which laid the foundation for robotic autonomous movement and intelligent navigation.Therefore,based on the research of SLAM theory,a set of Graph-SLAM system is built with lidar sensor to solve the problem of simultaneous localization and mapping for robot autonomous navigation in this thesis.First,the overall design of the Graph-SLAM system is completed.After analyzing the modeling methods and the adaptability to the environment with particle filter SLAM and Graph-SLAM,the more advantageous Graph-SLAM method is used to construct the system,and the hierarchical development framework of the Graph-SLAM system is designed.Then,the construction of the front end of the Graph-SLAM system is completed,and the back-end nonlinear optimization theory and loop closing detection method of the system is deeply studied.On the basis of graph modeling and lidar measurement modeling,the strategy of front-end information association is discussed.After realizing the information processing flow,the logical pseudo-code of the system front-end algorithm is given.Under the premise of Markov's hypothesis,the target optimization function of the system is derived,and the sparse feature of the system information matrix is used to simplify the solution process.In addition,a new loop closing detection method based on submap to submap matching is proposed for overcoming the current shortcomings of closed-loop detection about laser radar.Next,the map construction technology is researched and the robot autonomous navigation experiment based on Graph-SLAM system is conducted.After analyzing themainstream map models and the mapping methods,the active mapping method guided by RRT algorithm is used to construct the two-dimensional grid map.The global path planning algorithm and the obstacle avoidance control algorithm are added to the system for robot autonomous navigation test.The final experimental results show that the robot is performed perfectly about autonomous navigation in the unfamiliar environment under the lidar Graph-SLAM system designed in this thesis.Finally,the whole research works are summarized and the future research points are prospected in the thesis.