MM-ABC: Towards Generalist Mobile Manipulation via Seeing, Coordinating and Imagining

Hugging Face Daily Papers Papers
mobile-manipulation foundation-model robotics vision-language-model world-model attention-mechanism

Summary

MM-ABC提出一个面向通用移动操作的基础模型，通过稀疏多层VLM特征实现空间感知、以世界想象与几何意图为监督的训练分支，以及MM-APT中用掩码联合注意力协调操作与移动流。模型在5000+小时、17种本体的数据上预训练，在RoboCasa365、LIBERO等基准及真实移动操作任务上取得强性能。

Mobile manipulation extends robot interaction beyond a fixed kinematic workspace by making the reachable region itself controllable. This flexibility introduces two central challenges: spatially grounded perception under continuous ego-motion and coordinated control of heterogeneous arm and base actions. Existing approaches strengthen geometry through explicit 3D representations or predictive world models, and often decouple mobility and manipulation into separate action streams. We argue that effective mobile manipulation requires not only decoupling, but also representations that support efficient cross-stream collaboration. We present MM-ABC, a foundation model built around Seeing, Coordinating, and Imagining Arm-Base Collaboration. MM-ABC combines sparse multi-level VLM features for spatial perception; a training-only future branch that uses world imagination and geometric intent as extra supervision, strengthening perception and manipulation-intent prediction and improving the overall learning signal; and MM-APT, which coordinates separate manipulation and mobility streams through masked joint attention and clean-action x-prediction. In controlled ablations, replacing clean-action prediction with velocity prediction lowers success on RoboCasa365 composite-seen tasks from 32.8% to 29.2%, and removing future supervision or multilevel conditioning causes larger drops. We pretrain MM-ABC on 5,000+ hours of heterogeneous robot data spanning 400K+ episodes, 12 datasets, and 17 embodiments. Experiments cover EBench, RoboCasa365, ManiSkill-HAB, LIBERO, LIBERO-Plus, and real-world mobile manipulation. MM-ABC achieves 44.71% success on EBench, 61.2% on RoboCasa365, 99.1% on LIBERO, 82.8% on LIBERO-Plus without perturbation training, and 83% mean success on five real-world tasks.
Original Article
View Cached Full Text

Cached at: 10/06/26, 04:56 PM

Paper page - MM-ABC: Towards Generalist Mobile Manipulation via Seeing, Coordinating and Imagining

Source: https://huggingface.co/papers/2609.35652

Abstract

Mobilemanipulationextendsrobotinteractionbeyondafixedkinematicworkspacebymakingthereachableregionitselfcontrollable.Thisflexibilityintroducestwocentralchallenges:spatiallygroundedperceptionundercontinuousego-motionandcoordinatedcontrolofheterogeneousarmandbaseactions.Existingapproachesstrengthengeometrythroughexplicit3Drepresentationsorpredictiveworldmodels,andoftendecouplemobilityandmanipulationintoseparateactionstreams.Wearguethateffectivemobilemanipulationrequiresnotonlydecoupling,butalsorepresentationsthatsupportefficientcross-streamcollaboration.WepresentMM-ABC,afoundationmodelbuiltaroundSeeing,Coordinating,andImaginingArm-BaseCollaboration.MM-ABCcombinessparsemulti-levelVLMfeaturesforspatialperception;atraining-onlyfuturebranchthatusesworldimaginationandgeometricintentasextrasupervision,strengtheningperceptionandmanipulation-intentpredictionandimprovingtheoveralllearningsignal;andMM-APT,whichcoordinatesseparatemanipulationandmobilitystreamsthroughmaskedjointattentionandclean-actionx-prediction.Incontrolledablations,replacingclean-actionpredictionwithvelocitypredictionlowerssuccessonRoboCasa365composite-seentasksfrom32.8%to29.2%,andremovingfuturesupervisionormultilevelconditioningcauseslargerdrops.WepretrainMM-ABCon5,000+hoursofheterogeneousrobotdataspanning400K+episodes,12datasets,and17embodiments.ExperimentscoverEBench,RoboCasa365,ManiSkill-HAB,LIBERO,LIBERO-Plus,andreal-worldmobilemanipulation.MM-ABCachieves44.71%successonEBench,61.2%onRoboCasa365,99.1%onLIBERO,82.8%onLIBERO-Pluswithoutperturbationtraining,and83%meansuccessonfivereal-worldtasks.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2609\.35652

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.35652 in a model README.md to link it from this page.

Datasets citing this paper1

#### Kivy/MM-30 selfcollect_mobile_dualarm• Updatedabout 4 hours ago • 2.44k episodes • 1.38k • 1

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.35652 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

ABot-M0.5: Unified Mobility-and-Manipulation World Action Model

Hugging Face Daily Papers

ABot-M0.5 is a new World Action Model for mobile manipulation that improves performance through temporal granularity alignment, action space disentanglement, and train-test consistency, achieving state-of-the-art results on long-horizon and fine-grained manipulation benchmarks.

openbmb/MiniCPM-RobotManip

Hugging Face Models Trending

OpenBMB releases MiniCPM-RobotManip, a 1.5B vision-language-action model for on-device robot manipulation that outperforms larger models with efficient streaming inference and long-horizon visual memory.

ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models

Hugging Face Daily Papers

ACE-Brain-0.6 通过约150万条可验证视觉定位数据微调 Qwen3.5-9B，在七个公开定位基准中的四个上超越 GPT-6 Astra，展示了扩展多样化定位监督数据可显著提升开放多模态模型的空间感知与具身操作能力。

MotorMind: Scaffolding General Vision Language Models for Zero-Shot Robot Manipulation

Hugging Face Daily Papers

MotorMind introduces a harness that lets a general-purpose VLM control a robot directly by reasoning over observations and issuing mid-level actions with asynchronous feedback, achieving 66.7% success on LIBERO-PRO zero-shot (vs. up to 13.3% prior methods) and 95% on a real xArm6 robot without task-specific training, coding agents, or grounding tools.