Estimating human muscular fatigue in dynamic collaborative robotic tasks with learning-based models

Assessing human muscle fatigue is critical for optimizing performance in physical human–robot interaction (pHRI) tasks and mitigating safety risks for the human operator. This study presents a data-driven framework for estimating muscle fatigue in dynamic pHRI tasks using surface electromyography (sEMG) sensors attached to the human arm. Subject-specific machine learning (ML) regression models were developed to estimate fatigue levels during cyclic (i.e., repetitive) pHRI tasks. Specifically, Random Forest, XGBoost, and Linear Regression models were trained to estimate the fraction of cycles to fatigue (FCF) using three frequency-domain and one time-domain EMG features. Their performance was benchmarked against a convolutional neural network (CNN) that processes spectrogram representations of filtered EMG signals. Unlike most earlier data-driven approaches that primarily formulated fatigue estimation as a classification problem, our method models the continuous progression of fatigue through regression, enabling tracking of gradual physiological changes rather than discrete states, which is critical for timely intervention and adaptive control in dynamic pHRI tasks. Experiments were conducted with ten participants who interacted with a collaborative robot operated under an admittance controller, performing lateral (left-right) cyclic movements of the end effector until the onset of muscular fatigue. The results demonstrate that the root mean square error (RMSE) of FCF estimation across participants was 20.8 ± 4.3\%, 23.3 ± 3.8\%, 24.8 ± 4.5\%, and 26.9 ± 6.1\% for the CNN, Random Forest, XGBoost, and Linear Regression models, respectively. To examine cross-task generalization in this investigational study, additional experiments were performed with one participant who executed vertical (up–down) and circular repetitive movements. Models trained solely on the lateral-movement data were directly tested on these unseen tasks. The results indicate that the proposed ML/DL models are robust to variations in movement direction, arm kinematics, and muscle recruitment patterns, while the Linear Regression model performed poorly.