Adaptive Neural Network Impedance Control Of Robot And Human-Robot Physical Interaction

Hot topics in the study of robot have been gradually transferred to the "human-centered" collaborative robot.Human-robot collaboration has become a hotspot in the field of intelligent robots,which is widely applied in advanced manufacturing,health care and social services.Due to the inherent nonlinear characteristics of the robotic systems and the disturbance of the external environment,the model-based tracking control is hard to be applied in practical applications considering uncertainties in robotic dynamics.Moreover,because human is considered as an unknown time-varying dynamic interactive environment,traditional interaction control methods of robot based on known structure environment is difficult to be applied in human-robot interaction.Therefore,considering the uncertainties in robotic dynamic model and human behaviors,it is of great significance to propose an effective high-precision motion control method and an interaction control strategy with adaptability to unknown environment,for enhancing the safety,flexibility and intelligence in human-robot interaction.This paper is mainly aimed at the scenarios where human and robot complete tasks collaboratively through physical interaction.Focusing on uncertainties in robot dynamics and disturbances,state constraints,interaction force regulation,human motion intention estimation,human impedance learning,etc,we propose adaptive neural network and fuzzy neural network tracking control of robot,adaptive neural network impedance control of flexible joint robot,human-robot interaction control,Bayesian estimation method for human motion intention and human impedance,etc.In addition,experimental platforms are established to realize efficient human-robot collaboration,and the effectiveness and practicability of proposed control algorithms are verified.The following main research works are carried out:(1)Aiming at dealing with the unknown disturbance of external environment and the uncertainties in robot dynamic model,the adaptive neural network control strategy with disturbance observer of robot is proposed,and the full-state feedback and output feedback adaptive fuzzy neural network control strategy with output constraints of robot are studied,which improves the tracking accuracy in constrained space.(2)Aiming at the issue that it is difficult to adjust the flexibility for flexible joint robot,an adaptive neural network impedance control algorithm of flexible joint robot is proposed,which integrates the advantages of active compliance control and passive mechanical flexibility,solving the interaction problem when the dynamic model of flexible joint robot is unknown.Our proposed method enables the robot to adjust its flexibility and avoid damages caused by sudden collisions.(3)Aiming at the uncertainties in human behaviors in human-robot interaction,a human-robot interaction control strategy combining human motion intention estimation with impedance parameter learning is proposed,which can not only realize the motion constraint for safety,but also promote the interaction compliance.(4)Aiming at difficulties in human impedance learning and motion intention estimation for robot,an estimation method based on Bayesian theory is studied,an adaptive neural network impedance control strategy is proposed.Human impedance learning and motion intention estimation are integrated into the control framework to enhance robot learning ability of human behavior.(5)Aiming at the practical problems in human-robot cooperative tasks,the human and dual arm robot cooperative assembly platform and human-robot co-transporting platform are developed to complete several human-robot cooperative tasks,which further verifies the effectiveness and practicability of proposed control algorithms by platforms.The research enriches the current researches on human-robot interaction control of robot,further provides feasible solutions for solving the engineering problems existing in human-robot collaborative systems,and also provides a new research orientation for the research of human-robot physical interaction.