TU Delft Finds Personalized Exoskeleton Gait Offers No Comfort Boost

In a 10-person exoskeleton pilot, walking patterns fitted to each user scored no better than an averaged pattern. How long people had been in the device mattered more.

·
·
·
TU Delft Finds Personalized Exoskeleton Gait Offers No Comfort BoostPRO
Read2 min
TypePaper
TopicRobotics · Infra
  • A TU Delft-led pilot study found personalized exoskeleton gait patterns felt no better than standard ones.
  • Ten healthy adults walked with personalized, averaged and random patterns in a modified Lokomat with hip and pelvis actuation.
  • Personalization used small regressions on speed, height, weight, age and sex, trained on the Fukuchi dataset.
  • Trial order dominated the results: third trials were rated significantly more comfortable and natural than first trials.
  • Random patterns raised knee interaction forces by about 69 to 85 N. Personalized patterns increased reported physical strain.
  • Published in Frontiers in Robotics and AI. The authors say code and data will be posted on Zenodo.

A pilot study from TU Delft and partner institutions tested whether personalized gait trajectories improve the experience of walking in a rehabilitation exoskeleton. Ten healthy adults walked with three patterns: one generated from their body measurements, one based on an average gait, and one drawn from another person’s data. The study detected no significant effect of trajectory on comfort, enjoyment, naturalness, or overall preference. Ratings improved during later trials, suggesting that familiarization mattered more than the selected pattern.

The preprint has been published as a peer-reviewed journal article. The result gives exoskeleton developers a useful boundary: a personalization model adds little value when its trajectories barely differ from a population average and users are still adapting to the hardware.

Why stiff guidance needs a path

Trajectory-based lower-limb exoskeletons store a reference path for each joint across the gait cycle, then apply forces that keep the user’s limbs near those paths. Stiff guidance can reduce fall and injury risks for people with severe sensorimotor impairment or little control over the walking task.

Walking patterns vary with age, sex, body size, and speed, while reference trajectories commonly come from prerecorded population data. Researchers have therefore developed biomechanical and machine-learning methods that tailor those paths to individuals. Biomechanical models depend on assumptions about the body, while learning-based systems require enough representative training data to generalize.

Previous exoskeleton studies have rarely tested whether users perceive these personalized trajectories as more comfortable or natural. Subjective evaluations more often rely on broad usability questionnaires that do not isolate the walking pattern from hardware fit, controller stiffness, or adaptation.

Seven motors, 12 ways to move

The researchers used a heavily modified Lokomat, a commercial treadmill-based gait trainer developed with ETH Zurich and Hocoma. A standard Lokomat drives hip and knee motion in the sagittal plane, meaning the forward and backward movements visible from the side. The modified version also controls hip abduction and adduction, which move the leg sideways and back toward the body.

A spring-based pelvis module allows translation and rotation while actively driving side-to-side sway. The complete human-machine interface has 12 degrees of freedom, seven of which are motor-driven. Each hip uses two linear actuators: moving them together flexes or extends the hip, while moving them in opposite directions produces sideways motion.

Pro article

This story is for Pro members

You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.

Trending
  • No trending articles

Comments

avatar

Next Reads