Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2605.26284 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917533230039040 |
|---|---|
| author | Fujii, Koyo Figueredo, Luis Caleb-Solly, Praminda Boschi, Ivan Ida', Edoardo Carricato, Marco Magassouba, Aly |
| author_facet | Fujii, Koyo Figueredo, Luis Caleb-Solly, Praminda Boschi, Ivan Ida', Edoardo Carricato, Marco Magassouba, Aly |
| contents | Accurately estimating object mass and friction is fundamental to achieving reliable and adaptive robotic manipulation. Although interactive perception provides a powerful mechanism for inferring such properties, most existing approaches depend on specialized hardware such as force/torque sensors, tactile arrays, or multi-camera motion-capture systems, limiting scalability and deployment. This paper presents PhyPush, a physics-guided Transformer framework that estimates an object's mass and friction coefficient using only kinematically derived end-effector velocity from a single push. This typically requires data available on standard robotic arms. The model incorporates constraints from Newton's second law and the Coulomb friction model through a physics-guided loss, improving physical consistency and generalization to unseen objects and surfaces. Across diverse simulation and real-world setups, PhyPush consistently achieves more accurate mass and friction estimation in challenging out-of-domain conditions. In simulation, it reduces error by over 10% compared with a baseline that has privileged access to full force information, while in real-world experiments, it outperforms a data-driven loss approach. Overall, the results demonstrate that physics-guided learning can enable low-cost, sensor-efficient estimation of physical properties, relying solely on a single push and readily available kinematic data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_26284 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | PhyPush: One Push is All You Need for Sensorless Physical Property Estimation with Physics-Guided Transformers Fujii, Koyo Figueredo, Luis Caleb-Solly, Praminda Boschi, Ivan Ida', Edoardo Carricato, Marco Magassouba, Aly Robotics Accurately estimating object mass and friction is fundamental to achieving reliable and adaptive robotic manipulation. Although interactive perception provides a powerful mechanism for inferring such properties, most existing approaches depend on specialized hardware such as force/torque sensors, tactile arrays, or multi-camera motion-capture systems, limiting scalability and deployment. This paper presents PhyPush, a physics-guided Transformer framework that estimates an object's mass and friction coefficient using only kinematically derived end-effector velocity from a single push. This typically requires data available on standard robotic arms. The model incorporates constraints from Newton's second law and the Coulomb friction model through a physics-guided loss, improving physical consistency and generalization to unseen objects and surfaces. Across diverse simulation and real-world setups, PhyPush consistently achieves more accurate mass and friction estimation in challenging out-of-domain conditions. In simulation, it reduces error by over 10% compared with a baseline that has privileged access to full force information, while in real-world experiments, it outperforms a data-driven loss approach. Overall, the results demonstrate that physics-guided learning can enable low-cost, sensor-efficient estimation of physical properties, relying solely on a single push and readily available kinematic data. |
| title | PhyPush: One Push is All You Need for Sensorless Physical Property Estimation with Physics-Guided Transformers |
| topic | Robotics |
| url | https://arxiv.org/abs/2605.26284 |