Musculoskeletal Model-Based Adaptive Variable Impedance Control With Flexible Prescribed Performance for Rehabilitation Robots

Abstract

In rehabilitation robotics, both compliance and high-precision motion are critical to effective rehabilitation training. However, there is an inherent conflict between system compliance and control accuracy, which presents significant challenges in achieving optimal performance. To address this issue, this article proposes a surface electromyogram (sEMG)-driven musculoskeletal model-based adaptive variable impedance controller with flexible prescribed performance, ensuring a balance between compliance and precision motion in the human–robot interaction. We begin by formulating a constrained human–robot system and subsequently transform it into an unconstrained system using prescribed performance techniques. A novel impedance model is introduced to ensure system stability while maintaining prescribed performance. Furthermore, the controller integrates the Joint Strength Index, derived from an sEMG-driven musculoskeletal model, and incorporates a flexible prescribed performance function combined with adaptive stiffness and damping. The method supports both robot-dominant and human-dominant modes, and the seamless transition between them. Our findings show that lower human involvement increases system stiffness and narrows motion constraints, enabling high-precision motion. In contrast, greater human participation improves system compliance and broadens motion constraints, allowing for more freedom of movement. Finally, experiments were conducted on an upper-limb rehabilitation robot to validate the proposed method.