PocketDP3: Efficient Pocket-Scale 3D Visuomotor Policy

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Jinhao, Zhou, Zhexuan, Li, Huizhe, Lai, Yichen, Xia, Wenlong, Song, Haoming, Gong, Youmin, Mei, Jie
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912862556913664
author Zhang, Jinhao
Zhou, Zhexuan
Li, Huizhe
Lai, Yichen
Xia, Wenlong
Song, Haoming
Gong, Youmin
Mei, Jie
author_facet Zhang, Jinhao
Zhou, Zhexuan
Li, Huizhe
Lai, Yichen
Xia, Wenlong
Song, Haoming
Gong, Youmin
Mei, Jie
contents Recently, 3D vision-based diffusion policies have shown strong capability in learning complex robotic manipulation skills. However, a common architectural mismatch exists in these models: a tiny yet efficient point-cloud encoder is often paired with a massive decoder. Given a compact scene representation, we argue that this may lead to substantial parameter waste in the decoder. Motivated by this observation, we propose PocketDP3, a pocket-scale 3D diffusion policy that replaces the heavy conditional U-Net decoder used in prior methods with a lightweight Diffusion Mixer (DiM) built on MLP-Mixer blocks. This architecture enables efficient fusion across temporal and channel dimensions, significantly reducing model size. Notably, without any additional consistency distillation techniques, our method supports two-step inference without sacrificing performance, improving practicality for real-time deployment. Across three simulation benchmarks--RoboTwin2.0, Adroit, and MetaWorld--PocketDP3 achieves state-of-the-art performance with fewer than 1% of the parameters of prior methods, while also accelerating inference. Real-world experiments further demonstrate the practicality and transferability of our method in real-world settings. Code will be released.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22018
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PocketDP3: Efficient Pocket-Scale 3D Visuomotor Policy
Zhang, Jinhao
Zhou, Zhexuan
Li, Huizhe
Lai, Yichen
Xia, Wenlong
Song, Haoming
Gong, Youmin
Mei, Jie
Robotics
Recently, 3D vision-based diffusion policies have shown strong capability in learning complex robotic manipulation skills. However, a common architectural mismatch exists in these models: a tiny yet efficient point-cloud encoder is often paired with a massive decoder. Given a compact scene representation, we argue that this may lead to substantial parameter waste in the decoder. Motivated by this observation, we propose PocketDP3, a pocket-scale 3D diffusion policy that replaces the heavy conditional U-Net decoder used in prior methods with a lightweight Diffusion Mixer (DiM) built on MLP-Mixer blocks. This architecture enables efficient fusion across temporal and channel dimensions, significantly reducing model size. Notably, without any additional consistency distillation techniques, our method supports two-step inference without sacrificing performance, improving practicality for real-time deployment. Across three simulation benchmarks--RoboTwin2.0, Adroit, and MetaWorld--PocketDP3 achieves state-of-the-art performance with fewer than 1% of the parameters of prior methods, while also accelerating inference. Real-world experiments further demonstrate the practicality and transferability of our method in real-world settings. Code will be released.
title PocketDP3: Efficient Pocket-Scale 3D Visuomotor Policy
topic Robotics
url https://arxiv.org/abs/2601.22018