Improving Task-Agnostic Energy Shaping Control of Powered Exoskeletons with Task/Gait Classification

Area of Science:

• Robotics
• Biomechanics
• Machine Learning

Background:

• Task-agnostic control methods for powered exoskeletons offer versatility but often compromise performance for specific tasks or users.
• Existing energy shaping controllers provide stability but may not adapt optimally to individual user needs or varied daily activities.

Purpose of the Study:

• To develop and evaluate a novel control strategy for powered exoskeletons that integrates energy shaping with a machine learning classifier.
• To deliver optimal, personalized assistance across diverse tasks and users without explicit task detection, enhancing naturalistic movement.

Main Methods:

• A machine learning classifier was developed to detect transitions between multiple tasks and gait patterns.
• This classifier enables the selection of an optimized, task-agnostic controller from a weighted sum of pre-optimized energy-shaping controllers.
• An in-silico assessment was performed across various tasks, including incline walking, stair negotiation, and sit-to-stand transitions.

Main Results:

• The integrated control approach demonstrated superior performance compared to benchmark methods in 5-fold cross-validation.
• The method achieved 93.17 ± 7.39% cosine similarity and 77.92 ± 19.76% variance-accounted-for across diverse tasks and users.
• The results indicate significant adaptability in aligning exoskeleton assistance with human joint moments.

Conclusions:

• The proposed hybrid control strategy effectively enhances powered exoskeleton performance by adapting to individual users and tasks.
• This approach offers a promising solution for versatile and naturalistic human assistance in wearable robotic systems.
• Further research can explore real-world implementation and refinement of this adaptive control system.