Research On SLAM And Navigation Of Laser Vision Fusion Of Mobile Robot In Indoor Complex Environment

With the advancement of science and technology,and the economic development,mobile robots gradually play an increasingly important role in production and life,such as storage robots,medical robots,sweeping robots,etc.With the rapid development of computer hardware and related software,the research of mobile robot related algorithms has also entered a stage of rapid development.Among them,autonomous localization and mapping(SALM)and navigation of mobile robots are the focus and main content of research in the field of mobile robots.Many domestic and foreign scholars have studied robot autonomous positioning and map construction algorithms,thanks to the development of sensor technology.At present,the mainstream mobile robots use mostly wheel odometer,IMU,lidar(2D/3D),Camera(monocular/binocular/RGBD)etc.However,due to the principle of the sensor itself and the influence of environmental noise,for some complex scenes such as unstructured or environments with few geometric texture features,mobile robots cannot accurately and completely establish environmental maps and achieve navigation.In order to improve the autonomous positioning accuracy of the mobile robot and the completeness of the establishment of the environment map,this paper designs a multi-sensor fusion mobile robot SLAM and navigation system for complex indoor scenes and environments with few geometric features.Study the algorithm to build a complete environment map and improve the positioning accuracy.First of all,based on the Cartographer algorithm,study the internal and external reference calibration of the RGBD camera Kinectv2's point cloud information and 2D lidar,on this basis,the fusion of 3D point cloud and 2D point cloud information to establish a more comprehensive two-dimensional grid of environmental information map.Solve the incompleteness of the perception of environmental information when using only 2D lidar,and achieve a more complete expression of environmental information at low cost.Secondly,in order to improve the positioning accuracy of the mobile robot in the SLAM process and further improve the accuracy of mapping,for the environment with fewer geometric features such as long corridors,the positioning information provided by the wheel odometer,IMU,lidar,and RGBD cameras are separately Extraction,the use of extended Kalman filtering method,fusion of positioning information as the final SLAM positioning information,improve the positioning accuracy in the SLAM process,and then establish a more complete and accurate environment map,improve the accuracy of the SLAM process Sexuality and robustness.Next,the path planning under known map conditions is studied.In order to solve the positioning accuracy of mobile robots during navigation,based on the existing adaptive Monte Carlo(AMCL)algorithm,multi-sensor fusion is used to compensate for the original algorithm particle filtering.The disadvantage of large positioning error in an environment with few geometric features improves the positioning accuracy during navigation.Finally,based on the Robot Operating System(ROS),GAZEBO is used to build the robot model and environment model to verify the above algorithm.Experimental results show that the algorithm adopted in this paper can achieve a relatively complete and accurate expression of environmental information.It also improves the positioning accuracy during navigation using a single sensor.