Research On Iterative Learning Control Algorithms For Human-robot Collaboration

After years of development,the control algorithm of the collaborative robot is still based on learning from demonstration and impedance control,and in this case,it can't make better use of its advantages in cooperation with human.Therefore,this thesis focuses on the repetitive human-robot collaboration tasks,studies the collaborative robot control algorithm,the main control goal is to use the iterative learning control algorithm while working with human on collaboration tasks,to change the reference motion trajectory of the robot by recording the relevant position,force and speed information during the task.So as to adapt to human operation and sports habits,that can enhance human comfort in human-robot collaboration tasks and the accuracy of collaboration tasks.Based on the research of the kinematics and dynamics models of human-robot collaboration,aiming at the problem that the time domain iterative learning control algorithm cannot adapt to the uncertainty of human speed,this thesis proposes a spatial domain iterative learning control method based on contact force decomposition.This method makes use of the characteristic that the contact force in the normal direction is not affected by speed change.Through the analysis of the human contact force equation,with the goal of minimizing the contact force,the mathematical derivation and proof of the spatial domain iterative learning control algorithm is given.On the basis of learning the human space trajectory through the iterative learning control algorithm,in this thesis,the speed adaptation problem between the robot and the human is transformed into a position tracking problem,and the tracking error is minimized as the goal to predict the expected target position of the human at the next moment.And the relevant proof of convergence is given.This thesis simulates the algorithm based on repetitive human-robot collaboration tasks in MATLAB.It is verified by experiments that the iterative learning control algorithm can make the robot learn human movement habits and trajectories,which meets the expected goals of the subject.