Salvato in:
Dettagli Bibliografici
Autori principali: Zhou, Tianxing, Xue, Feiyang, Ye, Zhangchen, Yuan, Tianyuan, Zhao, Hang, Jiang, Tao
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2603.17720
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915872106348544
author Zhou, Tianxing
Xue, Feiyang
Ye, Zhangchen
Yuan, Tianyuan
Zhao, Hang
Jiang, Tao
author_facet Zhou, Tianxing
Xue, Feiyang
Ye, Zhangchen
Yuan, Tianyuan
Zhao, Hang
Jiang, Tao
contents Imitation learning is a prominent paradigm for robotic manipulation. However, existing visual imitation methods map 2D image observations directly to 3D action outputs, imposing a 2D-3D mismatch that hinders spatial reasoning and degrades robustness. We present VolumeDP, a policy architecture that restores spatial alignment by explicitly reasoning in 3D. VolumeDP first lifts image features into a Volumetric Representation via cross-attention. It then selects task-relevant voxels with a learnable module and converts them into a compact set of spatial tokens, markedly reducing computation while preserving action-critical geometry. Finally, a multi-token decoder conditions on the entire token set to predict actions, thereby avoiding lossy aggregation that collapses multiple spatial tokens into a single descriptor. VolumeDP achieves a state-of-the-art average success rate of 88.8% on the LIBERO simulation benchmark, outperforming the strongest baseline by a substantial 14.8% improvement. It also delivers large performance gains over prior methods on the ManiSkill and LIBERO-Plus benchmarks. Real-world experiments further demonstrate higher success rates and robust generalization to novel spatial layouts, camera viewpoints, and environment backgrounds. Code will be released.
format Preprint
id arxiv_https___arxiv_org_abs_2603_17720
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle VolumeDP: Modeling Volumetric Representation for Manipulation Policy Learning
Zhou, Tianxing
Xue, Feiyang
Ye, Zhangchen
Yuan, Tianyuan
Zhao, Hang
Jiang, Tao
Robotics
Imitation learning is a prominent paradigm for robotic manipulation. However, existing visual imitation methods map 2D image observations directly to 3D action outputs, imposing a 2D-3D mismatch that hinders spatial reasoning and degrades robustness. We present VolumeDP, a policy architecture that restores spatial alignment by explicitly reasoning in 3D. VolumeDP first lifts image features into a Volumetric Representation via cross-attention. It then selects task-relevant voxels with a learnable module and converts them into a compact set of spatial tokens, markedly reducing computation while preserving action-critical geometry. Finally, a multi-token decoder conditions on the entire token set to predict actions, thereby avoiding lossy aggregation that collapses multiple spatial tokens into a single descriptor. VolumeDP achieves a state-of-the-art average success rate of 88.8% on the LIBERO simulation benchmark, outperforming the strongest baseline by a substantial 14.8% improvement. It also delivers large performance gains over prior methods on the ManiSkill and LIBERO-Plus benchmarks. Real-world experiments further demonstrate higher success rates and robust generalization to novel spatial layouts, camera viewpoints, and environment backgrounds. Code will be released.
title VolumeDP: Modeling Volumetric Representation for Manipulation Policy Learning
topic Robotics
url https://arxiv.org/abs/2603.17720