Exo-plore: Exploring Exoskeleton Control space through Human-Aligned Simulation

Why we need physics simulation?

Previous methods requires demanding experiments involving hours of walking, paradoxically deterring those who would benefit most: people with mobility impairments.

Figures and videos from: [1] Gordon, Daniel FN, et al. "Human-in-the-loop optimization of exoskeleton assistance via online simulation of metabolic cost." IEEE Transactions on Robotics 38.3 (2022): 1410-1429.
[2] Ding, Ye, et al. "Human-in-the-loop optimization of hip assistance with a soft exosuit during walking." Science Robotics 3.15 (2018): eaar5438.

What is our optimization target?

Our goal is to find the optimal exoskeleton control parameters (κ, Δt) that minimize metabolic cost under torque assistance τ.

In our exoskeleton^[3], torque assistance (τ) is modulated by two control parameters: the gain (κ) scales the torque magnitude, while the delay (Δt) introduces temporal lag for the output (i.e., delayed-feedback control). The control signal is given by y(t) = sin(θ_r − θ_l), which encodes the relative motions of the two legs, where θ_r and θ_l denote the right and left hip joint angles, respectively.

[3] Lim, Bokman, et al. "Delayed output feedback control for gait assistance and resistance using a robotic exoskeleton." IEEE Robotics and Automation Letters 4.4 (2019): 3521-3528.

Framework Overview

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.

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 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.

Results

Gait Generation (Unassisted)

As a useful sanity check, Exo-plore generates realistic human gait patterns without assistive forces, with overall joint kinematics and variability closely resembling those of healthy walking.

Gait Generation (Assisted)

Exo-plore captures realistic hip kinematics and assistive dynamics across walking speeds, validating its effectiveness through consistent trends in both motion and power outputs.

Optimization Results

We demonstrate the applicability of Exo-plore by discovering and analyzing optimal exoskeleton control parameters for both able-bodied and disabled individuals.

Generalization to the pathologic gait

We evaluate Exo-plore across five musculoskeletal disorder types, revealing clear links between optimal exoskeleton assistance and the severity of each condition.

Below, we provide rich visualizations and detailed descriptions for each gait pathology.

* slowed x0.25 for detailed comparison.

Name	Cause	Description
Equinus	Contracture in Triceps surae	Walking on the toes due to persistent plantarflexion of the ankle. The heel does not touch the ground during gait.
Waddling	Weakness in Gluteus	Swaying of the trunk from side to side while walking due to weakness of the gluteal muscles, which fail to stabilize the pelvis. This gait is also referred to as a waddling or “duck-like” gait.
Crouch	Contracture in Psoas	Walking in a flexed, crouched posture, which maintains the hips in persistent flexion, often accompanied by knee flexion.
Calcaeous	Weakness in Triceps surae	Walking with the heel contacting the ground first due to weakness of the triceps surae muscles, leading to excessive dorsiflexion of the ankle.
Footdrop	Weakness in Tibialis anterior	Inability to dorsiflex the ankle during foot-off, resulting in the toes dragging on the ground. Patients often compensate with high-stepping gait.