Guardado en:
Detalles Bibliográficos
Autores principales: Tang, Yikai, Geng, Haoran, Jia, Jindou, Hu, Yuxuan, Zang, Sheng, Yang, Jianfei, Abbeel, Pieter, Malik, Jitendra
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2511.22445
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913177288048640
author Tang, Yikai
Geng, Haoran
Jia, Jindou
Hu, Yuxuan
Zang, Sheng
Yang, Jianfei
Abbeel, Pieter
Malik, Jitendra
author_facet Tang, Yikai
Geng, Haoran
Jia, Jindou
Hu, Yuxuan
Zang, Sheng
Yang, Jianfei
Abbeel, Pieter
Malik, Jitendra
contents Imitation learning has emerged as a crucial approach for acquiring visuomotor skills from demonstrations, where designing effective observation encoders is essential for policy generalization. However, existing methods tend to struggle once test-time conditions differ from the demonstrations, such as changes in lighting, texture, viewpoint, object placement, or object identity. To address this challenge, we propose DIffusion POlicy with compLementarity Encoders (DIPOLE), a visuomotor policy that learns to fuse complementary modalities through a training-time mechanism rather than a specialized fusion architecture. A modality-wise dropout masks one branch at each training step, encouraging each modality to remain individually informative. A lightweight cross-attention layer then exchanges complementary cues between the two. This design endows DIPOLE with five core strengths: stable high performance across diverse tasks, robustness to visual changes, spatial generalization at sub-centimeter precision, emergent capability beyond either modality, and zero-shot transfer to unseen objects. Across 18 simulated and 4 real-world tasks, DIPOLE outperforms six baselines by 39.1% on average, with gains of 41.5% under unseen visual distractors and 15.2% under randomized object placement.
format Preprint
id arxiv_https___arxiv_org_abs_2511_22445
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DIPOLE: Fusing Vision and Geometry for Robust Visuomotor Generalization
Tang, Yikai
Geng, Haoran
Jia, Jindou
Hu, Yuxuan
Zang, Sheng
Yang, Jianfei
Abbeel, Pieter
Malik, Jitendra
Robotics
Imitation learning has emerged as a crucial approach for acquiring visuomotor skills from demonstrations, where designing effective observation encoders is essential for policy generalization. However, existing methods tend to struggle once test-time conditions differ from the demonstrations, such as changes in lighting, texture, viewpoint, object placement, or object identity. To address this challenge, we propose DIffusion POlicy with compLementarity Encoders (DIPOLE), a visuomotor policy that learns to fuse complementary modalities through a training-time mechanism rather than a specialized fusion architecture. A modality-wise dropout masks one branch at each training step, encouraging each modality to remain individually informative. A lightweight cross-attention layer then exchanges complementary cues between the two. This design endows DIPOLE with five core strengths: stable high performance across diverse tasks, robustness to visual changes, spatial generalization at sub-centimeter precision, emergent capability beyond either modality, and zero-shot transfer to unseen objects. Across 18 simulated and 4 real-world tasks, DIPOLE outperforms six baselines by 39.1% on average, with gains of 41.5% under unseen visual distractors and 15.2% under randomized object placement.
title DIPOLE: Fusing Vision and Geometry for Robust Visuomotor Generalization
topic Robotics
url https://arxiv.org/abs/2511.22445