A Phase-Dependent and EMG-Driven Variable Stiffness Control Strategy for Upper Limb Rehabilitation Robot

Bilateral training is a widely adopted approach in upper limb rehabilitation due to its effectiveness and ease of implementation. Variable stiffness actuators enhance this approach by facilitating compliant human-machine interaction, ensuring both comfort and safety. However, optimizing stiffness levels for individual subjects to maximize training effectiveness remains a significant challenge. The difficulty of different task phases varies across subjects, necessitating customized robotic assistance. This paper proposes an adaptive stiffness modulation strategy that leverages the cyclic nature of rehabilitation tasks. An adaptive frequency oscillator is employed for real-time phase detection, while a tuning rule, based on motion tracking error and muscular activation, dynamically adjusts stiffness levels. Experimental validation with two participants demonstrated that the proposed strategy effectively promotes affected arm motion and tailors assistance to the user's specific needs.