Robotic Assistive Optimization and Control Using Neural Dynamics and Adaptive Neural Network.

Area of Science:

• Robotics
• Biomechanics
• Control Systems

Background:

• Human walking involves natural adaptation to diverse terrains.
• Soft exosuits offer promising assistive solutions but require sophisticated control for natural interaction.
• Existing controllers often struggle with real-time adaptation to human dynamics.

Purpose of the Study:

• To develop and validate a neural dynamics-based impedance optimization and trajectory adaptation approach for a dual-driven soft ankle exosuit.
• To enable the exosuit to learn and adapt to individual human ankle impedance and walking patterns in real-time.
• To enhance walking assistance by improving trajectory tracking and control stability.

Main Methods:

• A neural dynamics-based impedance optimization and trajectory adaptation strategy was implemented for a soft exosuit.
• An adaptive control framework integrated neural dynamics with adaptive laws for stable trajectory tracking.
• Lyapunov stability analysis confirmed the closed-loop system's uniform ultimate boundedness (UUB).

Main Results:

• The proposed method significantly reduced maximum trajectory tracking error to 0.016 rad, outperforming PID and ADRC controllers.
• Exosuit impedance parameters converged within 3 gait cycles across various terrains.
• Experimental validation on human subjects confirmed the controller's effectiveness in real-world walking scenarios.

Conclusions:

• The developed controller effectively integrates trajectory adaptation, force control, and impedance tuning for soft exosuits.
• The system provides a lightweight, wearability-optimized solution for enhanced walking assistance.
• This approach simulates human-like learning and adaptation in robotic assistive devices.