Geonho Leem1, Jaedong Lee2, Jehee Lee1, Seungmoon Song3, Jungdam Won1
1Seoul National University, 2Holiday Robotics, 3Northeastern University
ICLR 2026
[Project Page] [Paper] [Dataset]
Exo-plore finds optimal exoskeleton control parameters using musculoskeletal simulation only.
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 often unable 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.
See docs/installation.md for the full guide.
See docs/pipeline.md for the full pipeline with step-by-step commands.
- Train RL Policy (Gait Data Generator) — Train a musculoskeletal gait policy via PPO/Ray RLlib
- Rollout Gait Data Generator — Convert checkpoint to TorchScript, sample parameter space, and run C++ rollouts
- Train NN Surrogate — Learn a differentiable mapping from control parameters to metabolic cost
- Optimize Exoskeleton Parameters — Gradient-based optimization through the surrogate to find optimal (K, Delay) per walking speed
See docs/data.md for the complete data directory structure.
Configuration and resources (~105 MB) are under data/. Experiment data (~20 GB) are available for download from onedrive and should be placed under experiment_data/.
See docs/figures.md for instructions on reproducing all paper figures.
Geonho Leem — Seoul National University
This project is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Commercial use is NOT permitted without a separate license agreement. If you are interested in commercial use, please contact the author.