Real-Time Estimation of User Adaptation During Hip Exosuit-Assisted Walking Using Wearable Inertial Measurement Unit Data and Long Short-Term Memory Modeling

Author:  Cheonkyu Park, Alireza Nasizadeh, Kiho Lee, Gyeongmo Kim and Giuk Lee

Journal: Biomimetics

Published date: 2026.02

Link: https://www.mdpi.com/2313-7673/11/2/96

This study proposes a framework that uses wearable IMU sensor data and an LSTM model to estimate, in real time, how well a hip exosuit user has adapted to the device, without separate metabolic measurement equipment. In a six-day experiment with five participants, metabolic cost decreased by about 9.0% as participants adapted to the exosuit, and gait variability (variation in step frequency and hip joint angle) also decreased significantly, following a similar trend. The team trained an LSTM model on these gait-variability measures to predict each user's adaptation state; in cross-validation, it reached about 59.2% prediction accuracy within a ±10% error band of the metabolic-cost adaptation curve. This indicates that simple wearable sensors can track a user's physiological adaptation, which could support personalized control strategies that adjust robot assistance in real time.

Funding Information
*The Korea Medical Device Development Fund grant funded by the Korea government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) 
*The National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT)
*The Chung-Ang University Graduate Research Scholarship in 2025.

List