Research On Target Searching And Grasping Of Indoor Mobile Robot

With the continuous development of artificial intelligence and robot technology,mobile grasping robot has been widely used.Because of its complex application circumstance,it is difficult to perceive the environment accurately.Therefore,accurate recognition,precise positioning and efficient grasping are the key of mobile robot.In this paper,the key technologies of target recognition,positioning and grasping of mobile grasping robot are studied.Based on the robot operating system(ROS),an automatic control system of visual target searching and grasping is constructed.An efficient and accurate R-YOLO(ROBOTYOLO)network is proposed to optimize the robot target search process and is helpful to accurately complete the target grasping task.Finally,a robot platform is built for experimental test.The specific research work is as follows(1)According to the requirements of designing the robot,the hardware of the robot is designed and selected;the kinematics of the robot chassis is analyzed to obtain the direct mapping relationship between the robot odometer data and the rotation speed of Mecanum wheel;the robot control software system is established based on ROS,the mapping function package(Gmapping SLAM)and navigation function package(Navigation)were analyzed to implement the robot's automatic movement and obstacle avoidance functions.(2)On the basis of accurate and fast robot target recognition implemented by YOLO v4 algorithm,channel pruning algorithm is used for YOLO v4 to obtain a high-precision and real-time network which is called R-YOLO.The size of R-YOLO is 16% of YOLO v4;the detection efficiency of R-YOLO reaches 15 fps,and the detection accuracy and real-time performance meet the requirements of robot system.When the robot locates the target,the R-YOLO network is used to recognize the RGB image data collected by the RGB-D camera,and match with the point cloud data to calculate the spatial position of the target relative to the robot coordinate system.Thus the real-time and accurate target detection and positioning algorithm is realized.(3)By analyzing the kinematics principle of the manipulator,the coordinate system and D-H model of the four axis manipulator are established,the forward / inverse kinematics algorithm of the manipulator is studied,and the inverse kinematics solution process is optimized by geometric method.The experimental platform of mobile grasping robot is established,and the robot arm motion control and target grasping experiment are designed.The overall grabbing rate of the manipulator is 86.7%,and the grabbing rate of individual targets is more than 90%.The minimum grabbing time is 4.29 s,and the maximum grabbing time is 7.60 s.The robot can accomplish the target grasping task well in the real environment.