Sparse Bayesian Learning-Based Adaptive Impedance Control in Physical Human-Robot Interaction

For the sake of reducing human partner's effort (operating force and time) in human-robot interaction (HRI), it is of significant importance for robot to modify its impedance parameters dynamically based on human intention. Thus, in this paper, a data-driven adaptive impedance control (AIC) scheme is proposed, including a Sparse Bayesian learning-based (SBL) human intention predictor (HIP) and a variable impedance controller (VIC). And it works as follows: First, HIP is proposed to predict human partner's future intention by using necessary time-series data. Then, the predicted intention is used as an input to modulate impedance parameters by VIC. Thus, the dynamic characteristics of robot is suitable for operator's coming actions. Based on this, robot can adaptively comply to human partner better. The proposed method is verified by simulation on a 2 degrees of freedom (DOF) robot and experiments on a 6-DOF UR5 robot. Results reveal the feasibility and effectiveness of the proposed scheme in interaction process.