DSPv2: Improved Dense Policy for Effective and Generalizable Whole-body Mobile Manipulation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Su, Yue, Zhang, Chubin, Chen, Sijin, Tan, Liufan, Tang, Yansong, Wang, Jianan, Liu, Xihui
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916970624974848
author Su, Yue
Zhang, Chubin
Chen, Sijin
Tan, Liufan
Tang, Yansong
Wang, Jianan
Liu, Xihui
author_facet Su, Yue
Zhang, Chubin
Chen, Sijin
Tan, Liufan
Tang, Yansong
Wang, Jianan
Liu, Xihui
contents Learning whole-body mobile manipulation via imitation is essential for generalizing robotic skills to diverse environments and complex tasks. However, this goal is hindered by significant challenges, particularly in effectively processing complex observation, achieving robust generalization, and generating coherent actions. To address these issues, we propose DSPv2, a novel policy architecture. DSPv2 introduces an effective encoding scheme that aligns 3D spatial features with multi-view 2D semantic features. This fusion enables the policy to achieve broad generalization while retaining the fine-grained perception necessary for precise control. Furthermore, we extend the Dense Policy paradigm to the whole-body mobile manipulation domain, demonstrating its effectiveness in generating coherent and precise actions for the whole-body robotic platform. Extensive experiments show that our method significantly outperforms existing approaches in both task performance and generalization ability. Project page is available at: https://selen-suyue.github.io/DSPv2Net/.
format Preprint
id arxiv_https___arxiv_org_abs_2509_16063
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DSPv2: Improved Dense Policy for Effective and Generalizable Whole-body Mobile Manipulation
Su, Yue
Zhang, Chubin
Chen, Sijin
Tan, Liufan
Tang, Yansong
Wang, Jianan
Liu, Xihui
Robotics
Learning whole-body mobile manipulation via imitation is essential for generalizing robotic skills to diverse environments and complex tasks. However, this goal is hindered by significant challenges, particularly in effectively processing complex observation, achieving robust generalization, and generating coherent actions. To address these issues, we propose DSPv2, a novel policy architecture. DSPv2 introduces an effective encoding scheme that aligns 3D spatial features with multi-view 2D semantic features. This fusion enables the policy to achieve broad generalization while retaining the fine-grained perception necessary for precise control. Furthermore, we extend the Dense Policy paradigm to the whole-body mobile manipulation domain, demonstrating its effectiveness in generating coherent and precise actions for the whole-body robotic platform. Extensive experiments show that our method significantly outperforms existing approaches in both task performance and generalization ability. Project page is available at: https://selen-suyue.github.io/DSPv2Net/.
title DSPv2: Improved Dense Policy for Effective and Generalizable Whole-body Mobile Manipulation
topic Robotics
url https://arxiv.org/abs/2509.16063