Design of Robot Teleoperation System Based on Learning from Demonstration and Gaze Tracking
Aiming at the problem of traditional teleoperation system for medical robots,which lacks the cognition of task and the judgment of operator's subjectivity,a teleoperation system design strategy based on learning from demonstration and gaze tracking was proposed.Firstly,Gaussian mixture model(GMM)and Gaussian mixture regression(GMR)were applied to learn the robot's skills and realize the transfer of human-robot skills.Then,a virtual fixtures(VF)with artificial potential field method based on learning from demonstration was designed,and a force feedback algorithm was designed.Finally,a webcam was used to discriminate the operator's visual intention,and the intention recognition results and the artificial potential field method were applied to the virtual fixtures force feedback algorithm,and experiments were conducted.The research results show that the virtual fixtures force feedback teleoperation system based on visual intention discrimination can discriminate the operator's intention,effectively solve the problem that the operator's subjective intention was limited by the feedback force generated by the traditional virtual fixture,and enhance the operator's subjectivity in the process of teleoperation.It is an effective and practical force feedback control scheme in the teleoperation control system.The research results can provide a theoretical basis for the practical application of robot teleoperation system with the force-aware presence sensing motion in high-risk scenarios.
teleoperation systemlearning from demonstrationGMM(Gaussian Mixture Model)force feedbackintention recognition
万方数据
维普
