Research On Robotic Arm Control Of Personal Care Robot Based On Vision

With the aging of the population,personal care robot sits one of the hot spots.Personal care robots need to perform complex interactions with the environment,visual sensors to obtain information about the external environment,and then through the robot arm to interact with external objects,such as identifying an object and moving it to the target position.Therefore,the study of visual-based personal care robot robotic arm control has important theoretical significance and application value.The research object of this thesis is the hardware system composed of ASUS Xtion2 depth camera and robot arm of the personal care robot SEED.The software architecture is established in the open source Robot Operating System(ROS).Based on the DenavitHartenberg(D-H)coordinate system,the forward kinematics model of the robot arm is constructed.The inverse kinematics program of the robot arm is compiled in MATLAB,and the inverse kinematics problem of the robot arm is analyzed by numerical method.In the visual feedback part,project take the calibration target calibration method to calibrate the camera's internal parameters,external parameters and the hand eye relationship of the robotic arm system.In view of the problem that the characteristic extraction is not ideal due to the fact that the target object and the environment are not obvious in the actual work of the personal care robot,the method of AR tags attached to the object is effective.Aiming at the problem of depth information jitter,the instantaneous trigger method is used to collect depth image,and then the available real-time depth information is obtained.In order to verify the feasibility of the algorithm,the Unified Robot Description Format(URDF)model of the robot is used to realize visualization in the RVIZ platform,and the simulation and test environment of robotic arm path planning is built.In view of the problem of collision detection of the robot arm,the cylinder-enclosed box method is proposed,which simplifies the collision model and further adopts the segmented label sampling method,greatly reducing the amount of computation and shortening the collision detection time.Aiming at the defects of local minimum and oscillation in the traditional artificial potential field method,an improved method of adding anti-oscillation loop and sub target point is proposed.The simulation results of the traditional algorithm and the improved algorithm in MATLAB show that the improved algorithm not only solves the shortcomings of the traditional algorithm,but also makes more effective use of the global information.In this thesis,the depth camera positioning scheme,collision detection acceleration and the improvement of the robot arm path planning algorithm are studied,which lay the foundation for the actual hand-eye coordination control of the robot arm.