Human-Robot Skill Transfer And Control Based On Multi-Modal Information

With the deepening of the aging society and the increasing maturity of robot technology,people’s demand for the deeper integration of robots into human social production and life continues to improve.Operational skill transfer for complex force interaction tasks become one of the leading research directions in the field of human-robot skill transfer.By recording and analyzing the multi-modal skill data of skilled operators in the process of force interaction operation,a reasonable skill model is designed,which combines the control theory and robot perception ability to realize the robot autonomously executing complex force interaction tasks.So that robots can actually assist or even replace human beings to perform complex interactive tasks common in life.Starting from the mechanism of human-arm variable impedance operation in complex interactive tasks,this thesis studies the related problems of skill transfer and control in complex operation.The main contents include the following aspects: 1.Research on demonstration methods based on multi-modal information fusion;2.Research on the learning framework of complex interactive operation skills;3.Research on skill recurrence method based on interactive control.In the research on demonstration method based on multi-modal information fusion,considering that it is difficult for the traditional demonstration interface to demonstrate impedance parameter information intuitively,and the demonstration interface reacts too unnatural when interacting with people,this thesis designs a simplified humanarm impedance parameters model based on the antagonistic muscle co-contraction effect and human arm’s posture.And based on the model,it designs the variable impedance demonstration interface,realizing variable impedance dragging demonstration and teleimpedance demonstration methods etc..The arm impedance model parameters are obtained by parameter identification experiments combined with optimal algorithm,and the excellent demonstration performance and better human-robot cooperation effect of variable impedance demonstration method are verified by experiments.In the research on the learning framework of complex interactive operation skills,by combining human-arm operation mechanism and the design concept of dynamic system model,the interactive tasks for multiple operational actions are split;single dynamic system model is designed,then proposing dynamic system model for multiple operational action coupling,resulting the skill learning results obtaining more excellent human-robot collaborative performance.The Gauss model-based learning algorithm is used to train the interactive operation habits in the human-arm demonstration information.Based on the model,the interactive operation strategy is designed to help the robot controller achieve better operation performance.In the research of skill recurrence method based on interactive control.In order to solve the problems like big position error and switching slowly between different operations or even failure in the process of autonomous execution of multi-operation skill model,the motion state feedback variable impedance control strategy for switching freely between multi-operation tasks is proposed to achieve accurate execution of multioperation tasks and free switching between operation actions.Based on the operating mechanism and learning model of human arm,the model-based force feedback variable impedance controller is designed and the stability of the system is proved by energy tank theory.A model-based interactive force control algorithm is designed based on force feedback variable impedance controller.Through combing adaptive control method,the control accuracy of skill transfer framework for interaction forces is further improved.An experimental platform based on a 7-DOF single-armed robot is built,and the proposed complex operation skill transfer and control algorithms were verified by comparative experiments and demonstrated by comprehensive experiments.It verifies the feasibility and effectiveness of the proposed human arm’s multi-modal information demonstration method,complex interactive operation skill learning method,robots’ variable impedance operation skill control etc..Series of comparative tests under constant and variable impedance control verify the human-robot cooperation performance,high success rate and stability of control system of variable impedance human-robot interactive demonstration,interactive operation strategy based on learning models,impedance compliance switching strategy for multiple operational actions,variable impedance controller based on force feedback.Surface target interactive operational experiments verify the effectiveness,practicability and the good control performance for coupled dynamic system model,interactive operation strategy,interactive force control algorithm.Based on all the above work,this thesis has completed the design,construction and experimental verification of the variable impedance skill demonstration,learning and control framework for complex interactive tasks.Based on the robot’s imitation,learning and reproduction of the human operation skills,the system stability,the accuracy of interactive force control and the intelligence of autonomous execution of skills are further enhanced.