Research On The Mapping System Of Indoor Mobile Robot Based On Multi-sensor Fusion

In recent years,with the development of computer level and the continuous improvement of automation,robots have gradually entered people's lives.At present,mobile robots have been able to do more and more work for human beings,which has become a hot issue in current research,but also a key and difficult point.When a mobile robot enters a strange environment,without any information,it needs to build the scene map first.Only by completing the map construction in advance and making clear where it is,can we carry out the next work.This paper mainly focuses on the research of 3D map building system of indoor mobile robot,including:(1)Build the drawing system with the depth camera as the main sensor.Based on the feature point method,the front-end of the mapping system is designed,and the improved RANSAC method is designed to screen the matching point pairs,and then the motion is preliminarily estimated by ICP method.For the back-end construction,a key frame selection scheme is proposed to eliminate redundant key frames,which greatly improves the efficiency of the algorithm.After that,the g2o graph optimization is used to further optimize the pose estimation of the robot.For the loop detection module,K-means algorithm is used,and k-d tree data structure dictionary is used.Then TF-IDF weighted similarity function is designed to identify the loop,and the loop detection method with different detection schemes for different confidence intervals is proposed.All the above algorithms are verified by experiments,and the effect is good.(2)The movement of the wheeled robot is analyzed,the movement of the robot is modeled mathematically,and the systematic error of the odometer is eliminated through experiments,and the error of the gyroscope is corrected.(3)The sensor fusion scheme based on Kalman filter and extended Kalman filter is studied.Firstly,the odometer and gyroscope are fused based on Kalman filter,then the new fusion pose and the pose collected by depth camera are fused by extended Kalman filter,which greatly improves the accuracy and robustness of the algorithm.And the correctness of the algorithm is verified by comparing the real trajectory with the two sensor fusion and multi-sensor fusion pose estimation.(4)Data sets and experiments are used to verify the system.The experimental results show that the system is correct and robust.And the map processing module is designed.The threedimensional space map is processed into octree map which can be used for navigation and obstacle avoidance.