Muscle Fatigue-Aware Controller for a Semi-Rigid Knee Exoskeleton
Yifang Zhang, Jingcheng Jiang, Arash Ajoudani, Nikos Tsagarakis
AI summary
Problem
Traditional wearable exoskeletons rely on inconvenient EMG sensors or fixed assistance levels, making them impractical for long-term industrial use where dynamic fatigue assessment and continuous support are critical.
Approach
The method uses an offline-trained Gaussian Process Regression model to estimate muscle activation from standard joint kinematics and ground reaction forces, enabling online EMG-free fatigue monitoring that drives a predictive, optimization-based adaptive controller.
Key results
- Accurate online muscle activation estimation without EMG sensors
- Continuous fatigue monitoring across static and dynamic tasks
- Adaptive power output that conserves energy at low fatigue and increases support as fatigue rises
- Experimental validation demonstrating extended operational duration on a fixed battery
Why it matters
Provides a practical, sensor-light solution for industrial exoskeletons that dynamically balances worker support and device energy efficiency to prevent musculoskeletal injuries.
Abstract
Wearable assistive devices that monitor muscle fatigue reduce the risk of work-related musculoskeletal disorders, enhance rehabilitation outcomes, and extend operational time by optimizing the power consumption of the device. This work proposes a muscle fatigue-aware controller (MFAC) for a semi- rigid knee exoskeleton. During an offline calibration phase, we use Gaussian Process Regression (GPR) to model the relation- ship between muscle activation (measured via EMG) and the corresponding joint moment and angle, enabling fatigue state estimation for the controller. The trained model then approxi- mates muscle activation online using only joint states and moment derived from user’s kinematic data and ground reaction forces provided by the wearable device. The estimated muscle activation is used to assess the muscle fatigue state through a model-based fatigue evaluation module. Notably, EMG measurement is only required during the offline training in our approach, enabling EMG-free online estimation, which significantly enhances the feasibility for long-term mobile applications. Building on muscle fatigue and human–exoskeleton interaction models, we then developed an adaptive controller within a predictive control framework. The resulting optimization problem generates control signals that adjust assistance to reduce the fatigue progres- sion. Two experiments validate the EMG-free fatigue estimation method and the integrated MFAC, demonstrating accurate mus- cle activation estimation and effective adaptive assistance based on the estimated fatigue state. Analysis of actuator power output reveals adaptivity in which the controller conserves energy during low muscle fatigue and progressively improves power output with increasing fatigue, suggesting a longer duration of the device with a fixed battery capacity. Note to Practitioners—In industrial environments, many tasks still depend on repetitive human labor, often performed with limited rest and high physical demand. These conditions make workers vulnerable to muscle fatigue, a leading contributor to work-related musculoskeletal disorders. While wearable assistive devices are effective in mitigating fatigue, traditional solutions Received 19 May 2025; revised 25 August 2025 and 10 October 2025; accepted 14 November 2025. Date of publication 19 November 2025; date of current version 5 January 2026. This article was recommended for publication by Associate Editor X. Li and Editor Z. Li upon evaluation of the reviewers’ comments. This work was supported by European Union’s Horizon programmes under Grant 871237 (SOPHIA) and Grant 101070292 (HARIA). (Corresponding author: Yifang Zhang.) This work involved human subjects or animals in its research. Approval of all ethical and experimental procedures and protocols was granted by the Liguria Regional Ethics Committee: Studio IIT-HRII-SOPHIA–N. CER Liguria 554/2020. Yifang Zhang, Jingcheng Jiang, and Nikos G. Tsagarakis are with the Humanoids and Human Centered Mechatronics (HHCM) Research Line, Istituto Italiano Di Tecnologia (IIT), 16163 Genoa, Italy (e-mail: yifang.zhang@iit.it; jingcheng.jiang@iit.it; nikos.tsagarakis@iit.it). Arash Ajoudani is with the Human-Robot Interfaces and Interaction (HRI2) Research Line, Istituto Italiano Di Tecnologia (IIT), 16163 Genoa, Italy (e-mail: arash.ajoudani@iit.it). Digital Object Identifier 10.1109/TASE.2025.3634838 commonly rely on skin-contact electromyography sensors and deliver fixed assistance levels. Our proposed method avoids the need for such sensors during the operation period and enables adaptive assistance tailored to the user’s real-time fatigue condition, thereby ensuring smoother, safer, and more sustainable support for industrial workers. This paper introduces a muscle fatigue-aware controller for a knee exoskeleton that enables adaptive assistance based on the user’s fatigue state. One of the contributions is an offline training procedure that uses Gaussian Process Regression to capture the relationship between joint data and muscle activation, avoiding the need for complex biomechanical modeling. Once trained, the system can estimate muscle activation online using only the exoskeleton’s built-in motion and force sensors, eliminating the need for fragile and inconvenient electromyography electrodes during daily use. The estimated activation feeds into a model-based fatigue evaluation module, which enables continuous, online EMG-free fatigue monitoring and adaptive adjustment of assistance.Experiments demonstrate that this approach accurately estimates muscle activation, assesses fatigue reliably, and adjusts power output smoothly to balance assistance and device energy efficiency. While current results are limited to controlled laboratory testing with a single joint exoskeleton, the method has potential for broader, long-term use in industrial settings. Potential applications include industrial ergonomics, rehabilitation and sports training.