Model Predictive Control-Based Sensorless Physical Human-Robot Interaction of a Dual-Arm Social Robot

Recently, there has been increasing interest in physical human-robot interaction (pHRI) of social robots. This paper proposes a model predictive control (MPC)-based approach to realize various pHRI gestures of a social robot. For stable human-robot interaction, generating interactive motion while considering constraints such as workspace and joint velocity limitations is needed. In this study, MPC with adaptive weights is proposed to generate pHRI motion while accounting for these constraints. The weights of the MPC are adjusted to generate various social gestures based on the robot's posture and the external torques resulting from interactions with humans. The proposed methods are implemented on a dual-arm social robot. To realize the proposed method without additional sensors, a sensorless disturbance observer is employed to estimate external forces. Specifically, to mitigate unintended motions resulting from sudden disturbances during path generation, a regulated disturbance observer is proposed. To evaluate the effectiveness of the proposed methods, experiments were performed with a dual-arm social robot to produce pHRI motions like handshaking and hugging in different situation.