Reproduce “ExoTraj: A General Lower-limb Exoskeleton Assistance Policy for Complex Environments” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiao-Yin Liu, Guotao Li, Long Sun, Xu Liang, Zeng-Guang Hou
ExoTraj combines fast conditional flow matching with model predictive control to provide anticipatory hip assistance for lower-limb exoskeleton use across varied outdoor activities. Using only onboard IMUs and motor measurements, the system predicts 200 ms of future joint motion in about 5.09 ms, improves cross-subject prediction, and reduces metabolic rate by 14.7% on a complex outdoor route relative to zero-torque operation.
Key insights
- Predicting a short future trajectory allows the controller to anticipate human motion and optimize assistance over a horizon instead of reacting only to instantaneous tracking error.
- The reconstruction-augmented flow-matching model represents inter-subject trajectory variation while reducing inference time to roughly 5 ms; however, angular-velocity estimates remain less accurate near rapid motion peaks.
- MPC acts as a robustness and comfort layer: under injected Gaussian and impulse disturbances, the reported robustness index remained between 0.93 and 0.96 while limiting torque fluctuations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.