Open Source RoboticsRobots, software & learning

MobileManiBench

Software

Mobile-manipulation benchmark and task implementation built on IsaacLab, with external asset and data downloads. [1] [3]

Installation & first result →Official websiteGetting started

For
Researchers working on simulated mobile manipulation and robot-learning benchmarks [1]
Source rights
Named licenses are documented by part. Review licenses
Unknowns
operating system, compute requirements not fully established. Review gaps

Source researched · Not tested by us
Sources checked · Research details

Licenses & related hardware

Requirements · Research details and dates

On this page

Overview

MobileManiBench supplies a mobile-manipulation benchmark and task implementation built on Isaac Lab. Its overview distinguishes a G1 configuration with a parallel gripper from an XHand configuration with a dexterous hand, and lists opening, closing, pulling, pushing and picking skills. Those are benchmark configurations, not new manufacturing designs for the underlying robots. [1]

Is this for you?

Known limitations
  • Repository source includes inherited IsaacLab code and distinct unimanip task logic; scope is the benchmark, not a new hardware design.
  • External USD/object assets and datasets do not inherit the source-code BSD3 grant automatically.
  • No asset completeness, simulator execution, policy training, trajectory count or benchmark performance independently verified.

Editorial scope assessment from the cited implementation/documentation; no execution or compatibility test.

[1] [2] [3] [4] · Editorial assessment

What you need

Dependencies
  • README recommends Ubuntu22.04, driver550.x and CUDA12.x
  • Python3.10, torch2.5.1 CUDA12.1 wheel route and IsaacSim4.5.0
  • IsaacLab install plus editable unimanip package; external Assets/Assets.zip download

[1] [2] [3] [4]

Installation & first result

Open the upstream getting-started guide [1]

  • Follow the selected README environment route, inspect one unimanip task configuration and retrieve only its documented assets after reviewing their source terms.
  • Preserve exact source/dependency/model versions and evaluate the selected workflow before connecting physical hardware or reusing data.

Editorial scope assessment from the cited implementation/documentation; no execution or compatibility test.

[1] [2] [3] [4] · Editorial assessment

Installation

Follow the selected README environment route, inspect one unimanip task configuration and retrieve only its documented assets after reviewing their source terms.

Editorial scope assessment from the cited implementation/documentation; no execution or compatibility test.

[1] [2] [3] [4] · Editorial assessment

Hardware, models & data

Inputs and outputs

Simulated robot/object/scene states to task observations, rewards and generated manipulation trajectories.

Editorial scope assessment from the cited implementation/documentation; no execution or compatibility test.

[1] [3] · Editorial assessment

Project notes

Preserve the documented environment

The selected README recommends Ubuntu 22.04, a 550-series driver and CUDA 12.x. Its installation route uses Python 3.10, a torch 2.5.1 CUDA 12.1 wheel and Isaac Sim 4.5.0, followed by the Isaac Lab install and an editable unimanip package. Keep those source instructions with the selected repository revision rather than substituting a newer Isaac generation and assuming identical tasks. No dependency environment was installed here. [1]

This review selects repository revision 135466638d379c642f0c3f1742e6c20091b203c7. A concrete README example is the XHand, partnet, laptop setup using unimanip/configs/train/train_xhand_robot_open_best_0.yaml and the Isaac-XHand-Robot-Direct-v0 environment. Use that named setup to inspect the task configuration and required asset mappings rather than assuming a different object/robot pair has the same prerequisites. Asset completeness, GPU-memory capacity and storage requirements for that example remain unverified. [1] [4] [5]

The external asset step matters: the README directs users to download Assets/Assets.zip from arnoldland/MobileManiBench and extract it locally. Repository code alone is not a complete asset bundle. Select a named robot/object/task configuration and determine the needed USD files and data before attempting its training script. The inspected unimanip environment source imports Isaac Lab articulation, camera and scene interfaces, establishing task implementation but not a successfully loaded scene. [1] [3]

Separate code, assets and results

The inspected code and root notice use BSD-3-Clause with Isaac Lab attribution. That does not automatically license every external robot mesh, object model, scene or dataset. Review those assets' terms separately before copying or redistributing them. [2] [3]

Published task/trajectory counts and model results remain upstream reports. No external asset payload, simulator, policy training, generated trajectory or benchmark result was reproduced, and no physical robot was tested. [1]

What we checked

Source research only. Selected implementation/model text, repository tree and license declarations were inspected. No upstream compilation, dependency installation, simulator, network/USB/device integration, data collection, benchmark or physical robot test was performed. Source files are linked rather than redistributed; external services and paid products are not endorsed.

Not yet established

  • Requirements: operating system, host computer
  • Version-specific hardware compatibility (ecosystem tags alone do not establish it)

Updates

Upstream last updated
Not established
Latest release
Not established
Profile updated
Nov 1, 2026
Sources last checked
Nov 1, 2026
Next review
Dec 1, 2026
Reviewed revision
1354666

Source files and licenses

Distinct robot-facing simulation contribution with actual implementation/model content and a bounded documented next step. Review covers source artifacts, not operational performance or a complete tested deployment.

LayerAvailability / license
softwareavailable · BSD-3-Clause for inspected code; external model/data terms separate
Declaration for the scoped source implementation/documentation; dependencies, external services, model/data assets and inherited file-specific terms require separate checks. No blanket redistribution or operational clearance. [1] [2] [3]
documentationavailable · BSD-3-Clause for inspected code; external model/data terms separate
Declaration for the scoped source implementation/documentation; dependencies, external services, model/data assets and inherited file-specific terms require separate checks. No blanket redistribution or operational clearance. [1] [2] [3]

This profile is scoped to source implementation/model descriptions and an initial inspection workflow; no physical manufacturing design, pretrained checkpoint or dataset is selected for redistribution or tested deployment.

References

Each source links to the exact revision we read. “Used for” shows where this page relies on it.

Correction guidance · Discussion is not enabled · Profile format 0.7.0-draft.5

Change history