| With the development of robot technology,multi-robot cooperative operation provides diverse solutions for scenarios with multiple demand tasks in complex environments,and multirobot cooperative control requires stable and accurate multi-robot cooperative localization information.This thesis addresses the cooperative localization problem of multiple mobile robot platforms in indoor environments,and designs a distributed sensing cooperative localization method based on four omnidirectional mobile robot platforms under the framework of ROS robot operating system.The visual positional perception information is used to fuse the platform odometer state information,perform the relative positional filtering optimization based on distributed extended Kalman filter,and combine with LIDAR and IMU for laser SLAM global localization,and obtain the global positional state of the system as a whole through the global positional solution of some modules in the robot system.In this paper,four specific aspects of research are carried out as follows:(1)For the mobile robot platform state pose perception problem,we design a visual relative pose detection array based on monocular camera with full angle and no dead angle,and solve the relative pose between camera and target by observing AprilTag target on the robot platform,and then solve the relative observed pose between platforms through pose transformation;complete the connection between platform odometer and industrial control computer through CAN bus to obtain the feedback value of platform motion state in real time.(2)The mathematical modeling of different motion modes of the omnidirectional mobile robot platform is carried out,and the real-time motion state feedback of the platform is used as the system state quantity,and the visual relative perceived posture information is used as the system observation quantity,and the filter optimization system based on the extended Kalman filtering algorithm is constructed.The Kalman filters of the four platforms are coupled to build a distributed cooperative localization filtering system to maintain the bit-posture state information of the four platforms in real time,while adaptive optimization is performed for the visual perception errors in the distributed filtering system,and low-pass filters are constructed according to the relative observation distances to reduce the observation errors caused by increasing distances and make it more consistent with the actual perception model.(3)The LIDAR and IMU are built on specific modules in the four robotic platforms,and indoor laser SLAM mapping is performed based on the Cartographer laser SLAM algorithm,which is both a mapping and repositioning algorithm and can obtain the platform’s poses in the SLAM global map.Through coordinate transformation,the relative positional perception results of the remaining three platforms are mapped to the global coordinate system to realize the cooperative positioning of multiple robots in the global coordinate system.(4)Establish a multi-robot communication structure based on the ROS multi-robot communication protocol to provide lower communication latency and higher stability for the posture optimization system,while retaining a certain amount of expansion margin.(4)Build a simulation and physical platform to design experiments and verify the positioning accuracy and robustness of the distributed cooperative positioning sensing system.The proposed distributed cooperative positioning sensing system optimizes the odometer feedback information and the visual sensor sensing information,and realizes the global cooperative positioning on a low-cost platform by sensing the global posture of specific modules with SLAM algorithm,which effectively improves the positioning accuracy and robustness of the cooperative positioning system. |
