| With the widespread application of various types of indoor service robots in human life and production,Simultaneous Localization and Mapping(SLAM)technology for robots has gradually become a research hotspot.Due to the diversity and complexity of indoor environment,SLAM based on two-dimensional lidar can only obtain the obstacle information of installation height and the loop closure detection is difficult,so it cannot construct a high-precision environmental map with complete obstacle information.Therefore,aiming at the problems existing in the indoor environment of single two-dimensional lidar SLAM,this thesis proposes a multi-sensor fusion SLAM scheme.The main research contents are as follows:(1)Aiming at the problem that the Point-to-line Iterative Closest Point(PL-ICP)algorithm in two-dimensional lidar SLAM front-end motion estimation is sensitive to the initial value in the process of inter frame matching,the extended Kalman filter is used to fuse the data of Inertial Measurement Unit(IMU)and wheel odometer to provide the initial value for PL-ICP algorithm.The experimental results show that the positioning accuracy of the fusion odometer is improved by at least 33% compared with the simple wheel odometer,which solves the problem that PLICP algorithm requires high initial value and avoids the algorithm falling into local loop.(2)Aiming at the difficulty of loop closure detection of two-dimensional lidar SLAM,a synchronous motion vision sensor is used for loop closure detection,and then the loop closure information of vision detection is synchronously transmitted to the laser to help the laser realize the loop closure.The experimental results show that the robot’s position and posture are optimized and the root mean square error of the robot’s positioning is reduced by visual loop closure detection assisted by laser.(3)Aiming at the problem of incomplete obstacle information in the environment map constructed by two-dimensional lidar SLAM,the local two-dimensional grid map created by twodimensional lidar and the local two-dimensional grid map created by depth camera projection into pseudo two-dimensional laser data are fused.The experimental results show that the fusion compensates for the lack of vertical obstacle information in two-dimensional lidar,and constructs a more complete global environment map.(4)Finally,the thesis uses Handsfree robot platform with ubuntu 16.04 operating system and Robot Operating System(ROS)to implement a multi-sensor fusion indoor robot SLAM scheme,and conducts experiments in simulation and real environments respectively.The results show that the indoor robot SLAM scheme based on multi-sensor fusion improves the positioning accuracy of mobile robots and increases the integrity of building maps compared to a single twodimensional lidar SLAM scheme,which verifies the effectiveness of the fusion scheme. |
