Continual Online Personalization of Exoskeleton Control via Manifold-Aware Experience Replay

2026-06-15

AI Summary

This work addresses catastrophic forgetting in online personalized control of exoskeletons during task switching by proposing a manifold-aware experience replay framework. Without requiring explicit task labels, the method automatically distinguishes locomotion contexts through gait manifold identification and leverages an experience replay buffer to retain personalized policies from prior tasks, thereby unifying real-time adaptation with cross-task knowledge preservation. Integrating manifold learning, online adaptive control, and torque-phase tracking, the approach demonstrates substantial improvements in a simulated hemiparetic gait experiment: compared to a non-replay baseline, it achieves 40% higher torque tracking accuracy and 60% better gait phase tracking accuracy, significantly enhancing control robustness and personalization performance in multi-task scenarios.

Abstract

Personalizing exoskeleton control remains a critical challenge for clinical users with gait disabilities. Online adaptation (OA) offers an effective solution by adapting in real time to subject variability, device fit, and diverse locomotor tasks. However, OA involves a continual stream of user state data, which can lead to catastrophic forgetting of previously learned locomotor contexts. Here, we develop a manifold-aware experience replay-based online personalization framework designed to maintain user-specific representations across diverse tasks during OA of exoskeleton control. By replaying previously experienced tasks from a replay buffer, we preserve the personalized exoskeleton assistance across all learned tasks. Furthermore, we capture a gait manifold that distinguishes between different locomotor tasks, removing the need for explicit task labeling when selecting target replay bins. We evaluated our framework on emulated hemiplegic gait, which largely deviates from able-bodied patterns, across multiple forgetting scenarios with speed and incline transitions. Our manifold-aware replay framework achieved 40% and 60% improvements in torque and gait phase tracking accuracy, respectively, compared to a baseline framework without replay, which exhibited catastrophic forgetting during task transitions. This demonstrates that our proposed framework personalizes exoskeleton control in real time across diverse locomotor contexts in daily ambulation of clinical populations.

Problem

• continual learning
• catastrophic forgetting
• exoskeleton control
• online personalization
• gait adaptation

Innovation

• manifold-aware experience replay
• continual online personalization
• exoskeleton control
• catastrophic forgetting
• gait manifold