Research On Autonomous Navigation System Based On Multi-sensor Fusion SLAM

Autonomous navigation is the core topic of mobile robot research,and autonomous navigation based on SLAM(Simultaneous Localization And Mapping)technology is an important method to realize the autonomy of mobile robots.The fusion of multi-sensor data can combine the advantages of multiple sensors to help achieve robust perception.An important direction for the development of SLAM technology.Based on the SLAM technology of multisensor fusion,the thesis focuses on autonomous navigation in known maps and unknown environment application scenarios,and realizes the positioning,mapping and autonomous navigation in unknown environments.The specific research contents are as follows:Firstly,the two main problem in the autonomous navigation system-SLAM and path planning are modeled and solved respectively,and the mathematical model of sensor in the system is analyzed.Cartographer,a multi-sensor fusion SLAM algorithm for simultaneous localization and mapping in unknown environments,was studied.The IMU error in Cartographer will accumulate over time,which will lead to inaccurate back-end pose estimation.To solve this problem,an IMU and Li DAR are designed based on the theory of IMU preintegration.The tightly coupled front-end framework realizes joint nonlinear optimization of IMU pre-integration data and laser point cloud scanning matching data.Then,based on the map of grids constructed by SLAM technology,an autonomous navigation framework is studied to realize autonomous navigation of mobile robots,and the three modules of relocation,global path planning and local path planning in the framework are elaborated.AMCL(Adaptive Monte Carlo Localization)is adopted in the framework.The algorithm completes the relocation of the robot on a known map environment.As for global path planning,the effect of Dijkstra and A*with different heuristic functions in global path planning are compared and analyzed,and the A* algorithm with Manhattan distance is selected as the global path planning algorithm.As for local path planning,the DWA(Dynamic Window Approach)algorithm is selected,and in view of the defect that the classic DWA algorithm is easy to fall into local optimum,a DWA algorithm that integrates global path information is proposed as a local path planning algorithm.Finally,based on the research content,an autonomous navigation system based on multi-sensor fusion SLAM is designed,tested and analyzed.The results show that the optimized multi-sensor fusion SLAM algorithm has higher mapping accuracy,and the error in the constructed map is reduced by 0.35 cm per 100 cm compared with the previous one.The designed autonomous navigation system can complete the autonomous navigation task in the environment of known map,and the whole process of autonomous navigation is relatively stable.The designed autonomous navigation system can avoid obstacles appearing at any time in complex environment and has good real-time obstacle avoidance.