Learning-Based Torque Estimation for Harmonic Drive Actuators
Chun-Hung Huang, Chun Wei Chen, Chao-Chieh Lan
AI summary
Problem
Accurate torque estimation in harmonic drive actuators is hindered by nonlinear hysteresis and efficiency losses, making conventional calibration methods ineffective under varying loads and necessitating bulky external sensors.
Approach
A compliant helical tube is integrated into the drivetrain to generate encoder-measurable deformation features, which an LSTM network processes to model history-dependent dynamics and estimate output torque without additional hardware.
Key results
- Compliant tube design significantly improves estimation accuracy over stiffer or rigid alternatives
- Dual-encoder input (motor and wave generator angles) achieves the lowest error (3.96% RMSE)
- LSTM model successfully captures nonlinear hysteresis and generalizes across varying torques and impedance modes
- Online fine-tuning of the output layer enhances real-time estimation performance
Why it matters
This co-design strategy enables compact, reliable, and sensorless torque sensing for collaborative robots, reducing mechanical complexity while maintaining high accuracy under dynamic conditions.
Abstract
Accurate torque estimation in robotic actuators with harmonic drives is challenging due to nonlinear hysteresis and efficiency losses, often necessitating external torque sensors. This paper presents a learning-based torque estimation method that leverages encoder-derived features and mechanical compliance to enhance estimation accuracy without additional sensors. An actuator design incorporating a compliant helical tube provides deformation features that are effectively modeled using a Long Short-Term Memory (LSTM) network. Unlike conventional calibration or parametric approaches, the proposed framework captures nonlinear, history-dependent behaviors across varying operating conditions. Experimental evaluations demonstrate that compliant tubes significantly improve estimation accuracy compared with designs using stiffer or even rigid tubes, enabling more robust generalization under different torques, impedance modes, and stiffness levels. These results highlight the importance of co-designing actuator compliance and deep learning models to achieve reliable and compact torque estimation for harmonic drive actuators.