An Impedance Recognition Framework Based on Electromyogram for Physical Human–Robot Interaction

In physical human–robot interaction (pHRI), the interaction profiles, such as impedance and interaction force are greatly influenced by the operator's muscle activities, impedance and interaction force between the robot and the operator. Actually, parameters of interaction profiles are easy to be measured, such as position, velocity, acceleration, and muscle activities. However, the impedance cannot be directly measured. In some areas, it is difficult to capture the force information, especially where the force sensor is hard to be attached on the robots. In this sense, it is worth developing a feasible and simple solution to recognize the impedance parameters by exploring the potential relationship among the above mentioned interaction profiles. To this end, a framework of impedance recognition based on different time-based weight membership functions with broad learning system (TWMF-BLS) is developed for stable/unstable pHRI. Specifically, a linear weight membership function and a nonlinear weight membership function are proposed for stable and unstable pHRI by using the hybrid features for estimating the interaction force. And then the human arm impedance can be estimated without a biological model or a robot's model. Experimental results have demonstrated the feasibility and effectiveness of the proposed approach.