Exo-Plore: Exploring Exoskeleton Control Space through Human-aligned Simulation

TL;DR

Exo-plore addresses the challenge of optimizing exoskeleton control without burdensome human experiments by coupling a neuromechanical gait data generator with a surrogate optimizer trained from rich simulation data. It demonstrates sim-to-real alignment by reproducing human adaptation trends and then generalizes to five pathological gait types, revealing systematic relationships between pathology severity and optimal hip-assistance gains. The approach uses a composite reward and an intramuscular regularizer to promote physiologically plausible activations, and replaces traditional surrogate models with a large-data, gradient-friendly neural surrogate and Latin Hypercube sampling for robust optimization. Overall, Exo-plore offers a pathway to rapid, subject- and pathology-aware exoskeleton controller design with potential to accelerate clinical translation and reduce the dependence on extensive human testing.

Abstract

Exoskeletons show great promise for enhancing mobility, but providing appropriate assistance remains challenging due to the complexity of human adaptation to external forces. Current state-of-the-art approaches for optimizing exoskeleton controllers require extensive human experiments in which participants must walk for hours, creating a paradox: those who could benefit most from exoskeleton assistance, such as individuals with mobility impairments, are rarely able to participate in such demanding procedures. We present Exo-plore, a simulation framework that combines neuromechanical simulation with deep reinforcement learning to optimize hip exoskeleton assistance without requiring real human experiments. Exo-plore can (1) generate realistic gait data that captures human adaptation to assistive forces, (2) produce reliable optimization results despite the stochastic nature of human gait, and (3) generalize to pathological gaits, showing strong linear relationships between pathology severity and optimal assistance.

Figures (22)

• Figure 1: The Exo-plore framework. The generator produces gait trajectories under hip exoskeleton assistance using a musculoskeletal character with 164 muscles. Through our sim-to-real matching approach, these trajectories replicate the adaptation patterns and metabolic responses observed in real human experiments. The generated data are then used to train a surrogate network, which enables efficient and customizable optimization of exoskeleton control parameters.
• Figure 2: Comparison of joint kinematics over a gait cycle, with simulated results shown in red and human experimental data in gray.
• Figure 3: Comparison of simulated and experimental muscle activations.
• Figure 4: Walking speed vs (CoT, Step length, Step frequency)
• Figure 5: Comparison of gait kinematics and dynamics under exoskeleton assistance between human experimental results (top) and simulated results (bottom).