Transformer Observer-Based Contact Force Estimation for Quadruped Manipulators With Model Uncertainties
Fenghao Lin, Tianlin Zhang, Xiaogang Xiong, Yunjiang Lou
Abstract
Quadruped manipulators require precise detection of external forces to perform a series of force-related tasks during environmental interactions. However, these systems often lack tactile sensors on their body surfaces or force/torque sensors at critical joints. Conventional momentum based observer is widely used for external force estimation, but it relies heavily on an accurate dynamics model. This study introduces an external force estimation method for quadruped manipulators, utilizing a novel external force observer together with a Transformer network, that estimates external forces acting on the manipulator or the quadruped's base in the presence of model uncertainties. The observer leverages the robustness of a super-twisting algorithm (STA) based on the momentum model of quadruped manipulators, and compensates for the plantar force during dynamic gaits, reducing the influence on the estimation results. Model uncertainties are mitigated using Transformer, which predict the uncertainty-induced components in the observer estimation results. By means of residual, the proposed method estimates the purely applied external force with only proprioceptive sensors. Experimental validation across diverse scenarios confirms the effectiveness of this approach in facilitating safe and adaptive interactions between quadruped manipulators and external forces.
BibTeX
@inproceedings{ral2026_transformerobser,
title = {Transformer Observer-Based Contact Force Estimation for Quadruped Manipulators With Model Uncertainties},
author = {Fenghao Lin and Tianlin Zhang and Xiaogang Xiong and Yunjiang Lou},
booktitle = {RA-L 2026},
year = {2026}
}