Trajectory Planning Of Collaborative Robots Based On RGB-D Camera

Collaborative robots are characterized by man-machine integration,safety,ease of use,sensitivity,accuracy and flexibility.They are not only suitable for small-volume,multi-variety,and customized flexible manufacturing needs in the industrial field,but also have potential application prospects in the fields of service,medical and electric power.It has become an important direction for the development of the next generation of industrial robots.In the dynamic unstructured task environment,how to improve the environmental perception ability of the collaborative robot,so that it can adjust the trajectory dynamically according to the environmental change in real time,realize the autonomous obstacle avoidance and reach the target position,which is a key problem to be solved in the application research of the collaborative robot.Based on the key deployment project of the Chinese Academy of Sciences,"Key Technology and System Development of Next-Generation Industrial Robots"(Subject number:KGZD-EW-608-1),this dissertation studies the safety of the manipulator in unstructured environments.In this dissertation,the Kinect depth camera is used as the environment-aware sensor to study the environment-sensing and trajectory planning methods of the manipulator in the dynamic unstructured environment based on the UR manipulator.The main research work is as follows:The research and development history of the collaborative manipulator is expounded and analyzed.The research status of the environment perception and trajectory planning of the cooperative manipulator is analyzed,and the research significance of the trajectory planning of the cooperative manipulator in the dynamic unstructured environment is discussed.Taking the UR collaborative manipulator as the research object,the forward kinematics model and inverse kinematics model of the manipulator are established by D-H parameter method.The pose transformation relationship between the links of the manipulator is obtained.The kinematic model is simulated and verified by MATLAB.The vector product method is used to derive the Jacobian matrix of the manipulator,and the mapping relationship between the differential motion of the joint space and the differential motion of the operating space is established.The calibration principle of RGB-D depth camera is analyzed.Based on the hand-eye system composed of collaborative manipulator and depth camera,a hand-eye system calibration method based on least squares method is proposed.The mapping relationship between color image and depth image is analyzed.The fusion strategy of point cloud data is studied,and the 3D point cloud model of the manipulator workspace is established.Using the positive kinematics to divide the security domain of the workspace,combined with the LCCP algorithm,a point cloud image segmentation method of manipulator based on security domain is proposed,which realizes the rapid segmentation of the manipulator in the task environment.Compared with the workspace of the manipulator,the division of the security domain can reduce the data to be processed and ensure the safety of the manipulator itself.The position relationship between the manipulator and the obstacle is studied and analyzed.The octree structure is used to search and calculate the nearest point and the nearest distance between the manipulator and the obstacle,which lays the foundation for the dynamic trajectory planning of the manipulator.The overall obstacle avoidance strategy of the manipulator is analyzed,and the traditional local trajectory planning method is improved.A velocity-based trajectory planning algorithm based on null space is proposed to realize the real time obstacle avoidance of static and dynamic obstacles of the manipulator in the workspace.The method divides the target task and the obstacle avoidance task of the manipulator into different priorities,and the manipulator performs the target task under the premise of ensuring collision avoidance.If the manipulator cannot meet the obstacle avoidance task and the target task at the same time,the obstacle avoidance task is completed first,and the target task is temporarily abandoned.After the obstacle avoidance is completed,the target task is continued.The experimental platform of collaborative manipulator and RGB-D depth camera is built.The simulation and experimental research of the proposed method are carried out.The experimental data is analyzed,and the feasibility and effectiveness of the above method are further verified.