Knee Joint Torque Estimation via Fusion of Biomechanical and Data-Driven Models Using Exoskeleton Robot and sEMG

Accurate estimation of knee joint torque from surface electromyography (sEMG) is important for exoskeleton control, but remains challenging because muscle dynamics are nonlinear and vary across participants. This study proposes a hybrid framework that fuses a Hill-type biomechanical model with a data-driven CNN–LSTM model for sEMG-based knee torque estimation. The Hill-type model provides physiological interpretability through participant-level calibration of selected activation and scaling parameters, whereas the CNN–LSTM model captures nonlinear spatiotemporal features from sEMG signals. A residual fusion strategy was used, in which the biomechanical model served as the primary estimator and the data-driven model compensated for the residual error. Experiments with eight participants performing knee flexion–extension tasks showed that, relative to the model-derived reference torque, the fusion model reduced RMSE by 37.9% and 19.6% compared with the data-driven and biomechanical models, respectively. The proposed approach combines interpretability and adaptability, providing a feasible solution for personalized torque estimation in wearable robotic systems. Note to Practitioners—This work is motivated by the need for accurate and interpretable torque estimation in lower-limb exoskeleton control. In practical human–robot interaction, continuous estimation of movement intention is more suitable than predefined movement classification, but it is also more difficult. Compared with purely sensor-based motion recognition, sEMG provides earlier information about muscle activation and can therefore support more responsive assistance. The proposed method combines a Hill-type biomechanical model with a CNN–LSTM model. The biomechanical model provides a physiologically interpretable baseline estimate, while the data-driven model compensates for residual errors that are difficult to capture using simplified mechanics alone. This framework may be extended to other exoskeleton or prosthetic control scenarios, although joint-specific biomechanical models and sensor configurations would need to be redesigned for different applications.