Operability Improvement of Human-Robot Collaboration by Human-Adaptive Impedance Control Based on Human Arm Stiffness Estimation

Abstract

This paper proposes a variable impedance control for human-collaborative robots that can maintain the stability by adjusting parameters of an impedance model depending on stiffness of a human arm while keeping nimble so the human can maneuver an object easily. Specifically, first, the human mechanical impedance parameters (equivalent mass, viscosity, and stiffness of the arm) are identified based on a delta operator model. Based on the identification result, a recurrent neural network (RNN), which online estimates the human arm stiffness from rectified-and-integrated electromyogram (iEMG) signals is designed. Then, the estimate of the human arm stiffness, which is continuously changing and directly affects the contact stability, is used to adaptively change the parameters of the impedance model. By taking this approach, the proposed method aims to improve the operability of the human-robot collaborative system. Furthermore, changing the impedance parameters by additionally using speed of the manipulated object is also proposed to keep the operability nimble. To show the validity of the proposed method, experiments to evaluate the operability were conducted in which a variable-impedance controlled robot collaborated with a human to position an object. Experimental results showed that proposed method well adjust the impedance model depending on the human arm stiffness and can keep the system stable by predictively changing the impedance parameters to be more damped before unstable oscillation occurs. Furthermore, it has been confirmed that the 0 dB gain crossover frequency of the loop transfer function of the proposed method was 0.24 Hz while maintaining stability, the bandwidth of which was comparable to that of the light-fixed impedance case (0.22 Hz) that tends to become unstable as the arm stiffness increases. Also, by analyzing the human dynamics as a controller, it was found that the proposed variable impedance control allowed the human to conduct a task with less effort than that of a stable but heavy-fixed impedance case.