Open Source RoboticsRobots, software & learning

reBot Arm DLI Isaac Sim

Software
SimulationData collection

Isaac Sim 6.1 B601-RS teleoperation scenes that record leader-guided vial or stationery demonstrations as LeRobot v3 datasets. [1]

Installation & first result →Official websiteGetting started

For
Robot-learning developers with an NVIDIA RTX workstation; Builders with a reBot leader arm [1]
Hardware
Simulated reBot B601-RS, Physical reBot Arm 102 leader, NVIDIA RTX GPU [1]
Source rights
Named licenses are documented by part. Review licenses

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

Licenses & related hardware
Ecosystem
Isaac Sim, LeRobot

Requirements · Research details and dates

On this page

Overview

This project supplies an Isaac Sim 6.1 scene and control bridge for a simulated reBot B601-RS. A physical reBot Arm 102 leader sends joint input; the simulated arm follows it. The default task places three vials in a rack, while an optional scene places two pens and an eraser in a cup. The project records two camera views, joint states and actions as LeRobot Dataset v3 episodes. It is a data-collection workflow, not an autonomous policy or proof that a learned task transfers to a real B601. [1] [2] [3]

Is this for you?

Workflows
  • Teleoperate the B601-RS simulation
  • Record and inspect demonstrations
  • Replay actions and run a mock-leader regression

[1]

Known limitations
  • Simulation demonstrations are not evidence of real B601 task success.
  • Some bundled asset redistribution terms are unresolved; project root MIT covers only authored code/docs.
  • README's installation example still names the original yuyoujiang repository, so verify the intended Seeed checkout before following it.
  • Supplied notices do not clearly establish reuse terms for default SO-101-workshop vial/rack assets.

[1] [4]

What you need

Operating system
  • Ubuntu (version not fixed by the README)

[1]

Languages / runtimes
  • Python
  • Bash
  • USD scene descriptions

Editorial inference from the cited source; no runtime verification.

[2] [3] · Editorial assessment

Dependencies
  • Isaac Sim 6.1
  • Seeed LeRobot fork and leader/follower packages
  • git-lfs and ffmpeg
  • reBot Arm 102 leader for physical teleoperation
  • Python 3.12 and Seeed LeRobot installed with the dataset extra for recording/visualization

[1]

Host computer

Ubuntu workstation with NVIDIA RTX GPU; project recommends at least 12 GB VRAM and reports desktop freezes on 8 GB configurations.

[1]

Compute for the workload
  • NVIDIA RTX GPU with at least 12 GB VRAM recommended
  • Isaac Sim 6.1 on Ubuntu

[1]

Skills
  • Isaac Sim installation
  • LeRobot calibration and dataset inspection

Editorial inference from the cited source; no runtime verification.

[1] · Editorial assessment

Installation & first result

Open the upstream getting-started guide [1] · Project documentation [1]

  • Install Isaac Sim 6.1 and LeRobot dataset dependencies on Ubuntu.
  • Clone the Seeed repository and fetch its Git LFS assets.
  • Calibrate a reBot 102 leader over USB-UART and run the vial or stationery scene.
  • Record an episode and inspect camera/action synchronization in the LeRobot dataset.

Editorial inference from the cited source; no runtime verification.

[1] [2] [3] · Editorial assessment

Installation

Install Isaac Sim 6.1, Python 3.12 LeRobot packages and Git LFS assets, then run ./run.sh from the project checkout.

[1]

Hardware, models & data

Supported hardware
  • Simulated reBot B601-RS
  • Physical reBot Arm 102 leader
  • NVIDIA RTX GPU

[1]

Interfaces
  • USB-UART leader
  • localhost UDP:5005 leader bridge
  • LeRobot Dataset v3

[1] [2]

Inputs and outputs

Leader joint positions and scene state enter a simulated B601; wrist/overhead RGB streams, joint actions/states and task metadata are written as LeRobot v3 episodes.

[1] [3]

File formats
  • USD/USDZ assets and scenes
  • LeRobot Dataset v3

[1] [3]

Algorithms and models
  • Isaac Sim PhysX scene
  • Configurable domain randomization and grasp assist

[1]

Evaluation

Repository documents a mock-leader offline regression and episode replay, but OSR did not execute either or measure transfer to hardware.

[1]

Project notes

First useful run

The documented setup needs Ubuntu, Isaac Sim 6.1, Python 3.12, the Seeed LeRobot fork installed with its dataset extra, device packages, git-lfs, ffmpeg and a calibrated 102 leader over USB-UART. The authors recommend an RTX GPU with at least 12 GB of VRAM and report desktop freezes on 8 GB systems. After fetching the LFS assets, launch the vial scene with ./run.sh --variant vials, record an episode, then inspect whether the two camera streams and actions are synchronized. The README also describes replay and a mock-leader regression for checking the simulation path without the physical leader. We did not run either path. [1] [2] [3]

The repository README still shows a git clone command for the earlier yuyoujiang repository even though this reviewed project lives under Seeed-Projects. Pin the intended repository and revision before following its installation steps. The simulator's task success detector and randomized scenes are useful for dataset checks; they do not measure real-world pick-and-place success. [1]

Source and asset boundaries

The root MIT license covers project-authored code and documentation. THIRD_PARTY_NOTICES.md separates robot USD assets, a camera mount, Newton assets, Poly Haven stationery and a separately supplied workspace enclosure. Robot-asset redistribution needs upstream confirmation, and the workspace enclosure has no supplied license. The README says the default vials and rack came from an SO-101 workshop, but the supplied notices do not clearly establish their reuse terms either. Those asset sources and licenses need confirmation before copying the scene into another package. This review inspected the source and repository structure, including the LFS file list; it did not launch Isaac Sim, record a dataset or test a robot. [1] [4] [5]

What we checked

Source research and repository inspection only; no simulation launch, GPU performance measurement, dataset recording or physical robot test. No affiliate links or reused media.

Updates

Upstream last updated
Oct 14, 2026 [1]
Latest release
Not established
Profile updated
Oct 21, 2026
Sources last checked
Oct 21, 2026
Next review
Nov 21, 2026
Reviewed revision
865d8b3 · software repository

Source files and licenses

Distinct robot-specific Isaac Sim workflow and implementation; source is inspectable, but bundled asset permissions differ and some are unresolved. Describe source and asset boundaries without a physical transfer claim.

LayerAvailability / license
softwareavailable · MIT (project-authored source only)
Project-authored source is MIT. Robot, workspace and default vial/rack asset permissions are separate or unresolved; root MIT cannot be applied to them. [2] [3] [4] [5]
datasetsnot found · License not fully assessed
Code records datasets locally; no ready-made episode dataset was found in the inspected repository tree. [1]
documentationavailable · MIT (project-authored documentation only)
Linked third-party assets and media have separate terms. [1] [4] [5]

This hardware/software contribution does not publish this artifact type.

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

2026-10-21 — Created from Seeed reBot course discovery; pinned source research only.