Personalized Gait Patterns During Exoskeleton-Aided Training May Have Minimal Effect on User Experience. Insights from a Pilot Study

TL;DR

The study addresses the need for personalized gait patterns in exoskeleton-aided training by developing a data-driven, multi-planar personalization framework and a closed-form kinematic model for a treadmill-based exoskeleton. Using a regression-based approach trained on a normative database, personalized hip, knee, and pelvis trajectories are generated and enforced with a stiff PD controller, and compared against standardized and random patterns in a 10-participant pilot. Results show comparable trajectory accuracy across patterns, with no robust subjective advantages for personalization; instead, user adaptation over short trials improved comfort and naturalness irrespective of pattern. The findings highlight the importance of accounting for user adaptation in designing gait controllers and suggest that personalization may have limited short-term impact in unimpaired individuals, with future work needed in clinical populations and with longer familiarization.

Abstract

Robot-aided gait rehabilitation facilitates high-intensity and repeatable therapy. However, most exoskeletons rely on pre-recorded, non-personalized gait trajectories constrained to the sagittal plane, potentially limiting movement naturalness and user comfort. We present a data-driven gait personalization framework for an exoskeleton that supports multi-planar motion, including hip abduction/adduction and pelvic translation and rotation. Personalized trajectories to individual participants were generated using regression models trained on anthropometric, demographic, and walking speed data from a normative database. In a within-subject experiment involving ten unimpaired participants, these personalized trajectories were evaluated in regard to comfort, naturalness, and overall experience and compared against two standard patterns from the same database: one averaging all the trajectories, and one randomly selected. We did not find relevant differences across pattern conditions, despite all trajectories being executed with high accuracy thanks to a stiff position-derivative controller. We found, however, that pattern conditions in later trials were rated as more comfortable and natural than those in the first trial, suggesting that participants might have adapted to walking within the exoskeleton, regardless of the enforced gait pattern. Our findings highlight the importance of integrating subjective feedback when designing personalized gait controllers and accounting for user adaptation during experimentation.

Figures (4)

• Figure 1: (a) The new exoskeleton with a user secured, featuring an active body weight support (BWS) system with actuated lateral movement capability, a 6-DoF series elastic actuation (SEA) pelvis module with additional lateral translation actuator, two active 2-DoF closed-chain hip actuators, ball-screw-driven 1-DoF knee actuators, ankle/shank/thigh cuffs, and a safety harness for fall prevention. (b) Detail of the 6-DoF pelvis module. (c) Detail of the closed-chain mechanism for hip actuation.
• Figure 2: (a) Schematic of right leg kinematics where $O$ is the origin of the global frame, $A$ the origin of rotation axis of the hip actuators, $P_{int}$ and $P_{ext}$ are foreheads of the prismatic shafts, $E$ end-effector point of the parallel mechanism, $B$ head point of the back shaft, $M$ the projected point of $B$ on the thigh link, $H$ the hip joint, $P$ the pelvis plate and $F$ the fixed pelvis frame, and $K$ the knee joint. $l_m$ and $l_n$ are fixed segment lengths, with $l_n$ being the distance between the two non-actuated perpendicular revolute joints centered in $E$ and $B$, and $l_m$ the thigh link segment length (which can be modified based on the user's thigh length). (b) Top view of the hip actuators with parallel mechanism in the actuator plane; $l_c$ is the distance between the two parallel motors' linear shafts, $d$ is half the length of the linear actuators, and $P_{ext}$ and $P_{int}$ represent how far the motor shafts travel. $\theta _{A}$, $\theta _{B}$, and $\theta _{E}$ are unknowns.
• Figure 3: Top view of the right leg's closed-chain mechanism with the pelvis fixed in rotation, shown in different configurations: a) Neutral, b) Hip flexion/extension motion, and c) Hip ab-/adduction motion. Orange and purple arrows represent the movement of the linear actuators. Red and green arrows represent the resulting motion at the thigh.
• Figure 4: Exemplary right leg trajectories of a participant are displayed at walking speed levels 1, 2, and 3, together with their extracted key events. These key events are the same as in Koopman's study koopman2014, except for lateral pelvis movement (not present in their study), for which extreme position and velocity values were chosen. The key events for the left and right joint trajectories were extracted separately. By default, heel strike timing is set to zero (% of the gait cycle), as are the minimum and maximum values for joint position and velocity. These constraints are detailed in the key event legend.