Researchers have developed an active adaptive human-robot interaction control method for robot-assisted gait training, aiming to improve compliance and safety in lower-limb exoskeletons. The approach integrates three key components: active gait trajectory planning driven by surface electromyography (sEMG) signals, stiffness adaptive control using fuzzy logic, and gait replanning based on model predictive control (MPC).
Key takeaways
- Gait trajectory is parameterized using a 5th-order finite Fourier series, with cadence and stride adjusted online based on muscle activation derived from sEMG, enabling intention-driven training.
- A fuzzy logic controller adjusts joint equivalent stiffness in real time, using muscle activation and position deviation as inputs, to balance assistive force and stability.
- Dynamic balance and trajectory stability indices are defined; MPC replans gait trajectories online to prevent falls and maintain human-robot system stability.
- Simulations and experiments on a self-developed exoskeleton with active hip, knee, and ankle joints validated the method's effectiveness.
This work addresses limitations of existing methods that rely on kinematic errors or ground reaction forces, which may not accurately reflect patient intent. By leveraging sEMG for intent detection and adaptive control, the approach enhances patient engagement and interaction compliance, potentially improving rehabilitation outcomes for individuals with gait impairments.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-03-25 · “面向机器人辅助步态训练的主动自适应人机交互控制”