Research On Key Technologies Of Indoor Mobile Robot Based On Lidar

Since the 21 st century,with the prominence of social aging and the increase of human working-cost,the demand for robots has made the research on key technologies of robots mature and tend to be intelligent.Indoor mobile robots are important direction of intelligent mobile robots,and environmental map building,autonomous navigation and pedestrian detection,tracking and following are the key and core functions of indoor mobile robots.Therefore,it is of great significance to study them.Aiming at the research of indoor mobile robots,this research presents the overall design scheme of the software as well as hardware system and studies the indoor mobile robots based on two-dimensional lidar from three aspects which are: environment map building,autonomous navigation and target pedestrian following.Finally,the algorithm is designed and implemented.The overall work of this research includes four aspects:1)An odometer based on extended Kalman filter is proposed to fuse multi-sensor data in order to resolve the low accuracy issue of odometer which is based on single sensor.Two kinds of SLAM(Simultaneous Localization And Mapping)algorithms based on Rao-Blackwelllized particle filter and graph-based optimization principle are analyzed and studied.And improved Rao-Blackwelllized particle filter SLAM algorithm based on multi-sensor motion model is proposed.Then,the improved SLAM proposed by this work and Cartographer are compared and analyzed in the environment of crowded office and long corridor.Finally,the improved SLAM algorithm proposed is selected to construct the environmental map.2)The autonomous navigation of indoor mobile robots is studied from four aspects that are: cost map generation,self-localization,global path planning and local path planning.In the aspect of global path planning,two kinds of algorithms based on sampling and graph are compared and analyzed,and A* algotithm is selected according to simulation results.In the aspect of local path planning,two kinds of algorithm based on sampling and optimization are compared and analyzed,and timed elastic band algorithm is selected according to simulation results.3)Deformable parts model with LSVM pedestrian detection and kernelized correlation filters target pedestrian tracking schemes are selected after going through the research as well as analysis of common target detection features and target tracking algorithm.As,monocular camera can not obtain the distance between the pedestrians and the robot.Therefore,a method is proposed to calibrate the camera internal parameters first.Then,use the chessboard as the intermediate coordinate system to obtain the transformation from the lidar coordinate system to the camera coordinate system and fuse the data of two systems through the clamp constraint.Finally,the complete process of target pedestrian following is given.4)According to the hardware platform selected in this research,the key algorithms are implemented and integrated.The feasibility of the proposed method is verified by systematic experiments from three aforementioned aspects that are: construction of environment map,autonomous navigation,target pedestrian following.