UKF-Based Sensor Fusion for Joint-Torque Sensorless Humanoid Robots

TL;DR

The paper addresses sensorless torque control for humanoid robots by estimating joint torques from multimodal sensor data using an Unscented Kalman Filter-based sensor fusion framework. It fuses encoder, force/torque, inertial, and motor-current measurements to infer $\tau_j$ while compensating for external contacts and friction from harmonic drives, integrating a friction model $\tau_F = k_0 \tanh(k_1 \dot s) + k_2 \dot s$ and $\tau_j = r\tau_m - \tau_F$. The estimator operates within a two-layer torque-control architecture and is validated on the ergoCub humanoid, showing improved torque-tracking accuracy with RMS errors in the range $0.05$–$2.5\,$Nm compared to RNEA-based approaches. This approach enables robust, sensorless torque control in humanoid robots and supports flexible sensor configurations with external-contact robustness.

Abstract

This paper proposes a novel sensor fusion based on Unscented Kalman Filtering for the online estimation of joint-torques of humanoid robots without joint-torque sensors. At the feature level, the proposed approach considers multimodal measurements (e.g. currents, accelerations, etc.) and non-directly measurable effects, such as external contacts, thus leading to joint torques readily usable in control architectures for human-robot interaction. The proposed sensor fusion can also integrate distributed, non-collocated force/torque sensors, thus being a flexible framework with respect to the underlying robot sensor suit. To validate the approach, we show how the proposed sensor fusion can be integrated into a twolevel torque control architecture aiming at task-space torquecontrol. The performances of the proposed approach are shown through extensive tests on the new humanoid robot ergoCub, currently being developed at Istituto Italiano di Tecnologia. We also compare our strategy with the existing state-of-theart approach based on the recursive Newton-Euler algorithm. Results demonstrate that our method achieves low root mean square errors in torque tracking, ranging from 0.05 Nm to 2.5 Nm, even in the presence of external contacts.

Figures (2)

• Figure 1: On the left The ergoCub humanoid robot in transparency with FT sensors highlighted in green. On the right ergoCub being pushed on the torso during a closed-loop torque controller test.
• Figure 2: Block diagram of the controller architecture implemented on the ergoCub humanoid robot.