Research On Human Upper Limb Tracking Prediction And Robot Obstacle Avoidance Algorithm

The application of robot industry and related technologies is an important support for modern manufacturing.With the intelligent development of manufacturing industry,collaborative robot and human-robot collaboration technology have used widely.In the human-robot collaboration scene,there are uncertain factors such as dynamic obstacles dominated by human upper limb,which puts forward a higher test for human-robot safety strategy and robotic real-time path planning technology.Facing the human-robot collaboration assembly and other scenes,this paper studies the tracking and prediction of human upper limb position and the real-time obstacle avoidance planning of manipulator.Firstly,this paper introduces the structural parameters of the lightweight manipulator,established the linkage coordinate system according to the improved DH parameter method,and carried out kinematic modeling and verification.The singular point of the manipulator was analyzed,and the workspace of the manipulator was drawn by Monte Carlo method based on random sampling.Built a vision platform to study and analyze the relevant theories of vision sensor,the camera calibration and hand-eye calibration of infrared camera were carried out.Secondly,in view of the redundancy and error of the joint data collected by the camera,the data was grouped and sent.After screening the amount of data,the position can still be correctly expressed and the data fluctuation can be reduced;a coarse error elimination strategy based on the distance between joints was designed and verified.This strategy can filter and eliminate the coarse error and abnormal value in the collected data.A tracking and prediction method based on historical trajectory was proposed.The historical trajectory of joint motion is used to predict the next time node position by calculating the derivative speed by difference method,and it is corrected and supplemented when there are gross errors and abnormalities in the data collected by the camera.Based on the idea of OBB bounding box,a human-robot real-time collision detection model was established,and a calculation method of the actual minimum distance between the manipulator envelope and the obstacle was designed according to the distance relationship between spatial points and lines.Then,the mainstream path planning algorithms were studied and analyzed,and a path planning algorithm of manipulator in 3D space based on velocity potential field was proposed.The algorithm introduces a new attraction velocity function to solve the problem of flexible impact at the beginning and end position,and ensure the travel efficiency.A buffer zone was set for the repulsion velocity to reduce the sudden change of velocity in and out of the range of repulsion potential field.At the same time,a method of establishing virtual target points was proposed,and the trajectory was smoothed by fifth polynomial programming to optimize the path.Finally,an obstacle avoidance experimental platform was built,and the real-time obstacle avoidance experiment when the human upper limb is close to the manipulator was carried out based on the data collected by the vision sensor,and the proposed algorithm was verified.The results show that the running path of the manipulator under the proposed algorithm is relatively smooth,which can realize the simultaneous planning of path and speed,reduce the process of speed re planning,and meet the requirements of real-time obstacle avoidance of the manipulator.