SO-101 simulation: free URDF, MuJoCo and Isaac Sim models

SO-101

The SO-101 has the most sim support of any arm on this site. The maker ships URDF and MuJoCo files, and NVIDIA ships a USD model inside Isaac Sim.

• SO-ARM100 repo: Simulation/SO101

  The official files. There are new- and old-calibration versions and one with the wrist camera mount. Match the calibration you used on your real arm.

  URDF, MJCF, MuJoCo, or anything that loads URDF, by The Robot Studio

• MuJoCo Menagerie: robotstudio_so101

  A cleaned-up copy of the official new-calibration MJCF.

  MJCF, MuJoCo 3.1.3+, by Google DeepMind (MuJoCo Menagerie)

• Isaac Sim robot library: so101_new_calib

  Already in Isaac Sim at RobotStudio/so101_new_calib/so101_new_calib.usd. Apache 2.0.

  USD, Isaac Sim, Isaac Lab, by NVIDIA

• Train an SO-101 from sim to real (course)

  A free self-paced course. You teleop in sim, collect data, train a GR00T model and run it on the real arm. The workshop code was tested on RTX 5090 and RTX 6000-class GPUs.

  USD, Isaac Lab, by NVIDIA

• Sim-to-Real-SO-101-Workshop repo

  The code for the course: a vials-to-rack task, domain randomization and eval scripts.

  USD, Isaac Lab (Docker), by NVIDIA (isaac-sim)

• LeIsaac

  Drive the sim arm with your real SO-101 leader arm and record LeRobot datasets. LeRobot lists it in its EnvHub.

  USD, Isaac Lab, by Lightwheel

Builds on this site: Build the $100-class SO-101 open-source robot arm · Teach an SO-101 arm to pick and place with LeRobot · Run SmolVLA on an SO-101 so it follows plain-English tasks · Let a coding agent drive an SO-101 to paint or draw · Livestream and remote-control your SO-101 arm from a web app · Use a folding phone as a controller for an SO-101 gripper

SO-100

• SO-ARM100 repo: Simulation/SO100

  The official URDF with meshes.

  URDF, Anything that loads URDF, by The Robot Studio

• MuJoCo Menagerie: trs_so_arm100

  A simplified model made from the official URDF.

  MJCF, MuJoCo 3.1.6+, by Google DeepMind (MuJoCo Menagerie)

• Isaac Sim robot library: so100

  In Isaac Sim at RobotStudio/so100/so100.usd. Apache 2.0.

  USD, Isaac Sim, Isaac Lab, by NVIDIA

The build on this site: Build the $100-class SO-101 open-source robot arm

LeKiwi

• LeKiwi repo: URDF

  The official URDF with meshes for the base and arm. Joint names are listed in JOINT_NAMES.md.

  URDF, Anything that loads URDF, by SIGRobotics UIUC

• LeIsaac

  Added LeKiwi teleop and a loft scene in Dec 2025.

  USD, Isaac Lab, by Lightwheel

The build on this site: Put an SO-101 arm on wheels with LeKiwi

XLeRobot

• XLeRobot repo: simulation

  Official sim folder with URDFs for ManiSkill and a MuJoCo scene. The docs have a step-by-step install guide.

  URDF, MJCF, ManiSkill, MuJoCo, by Vector Wang

• MuJoCo-GS-Web demo

  Try XLeRobot in a web page, MuJoCo plus Gaussian-splat scenes. Nothing to install.

  MJCF (in the browser), MuJoCo in the browser, by Vector Wang

The build on this site: Build XLeRobot, a dual-arm home robot from about $660

Koch v1.1

• MuJoCo Menagerie: low_cost_robot_arm

  Menagerie's model of Koch v1.1, made from a community URDF. The koch-v1-1 repo itself has no sim files.

  MJCF, MuJoCo 3.1.6+, by Google DeepMind (MuJoCo Menagerie)

The build on this site: Build the Koch v1.1 Dynamixel arm for imitation learning

Original low-cost robot arm

• low_cost_robot repo: simulation

  In the original repo. teleoperate_simulated_robot.py lets your real leader arm drive the sim arm.

  URDF, MJCF, MuJoCo, by Alexander Koch

The build on this site: The original $250 low-cost robot arm with a leader arm

ALOHA

• gym-aloha

  Two tasks: hand a cube from one arm to the other, and insert a peg into a socket in mid-air. pip install gym-aloha.

  MJCF, MuJoCo (Gymnasium), by Hugging Face

• MuJoCo Menagerie: aloha

  This one is ALOHA 2, with the newer gripper, built from the ViperX 300 model.

  MJCF, MuJoCo 3.1.1+, by Google DeepMind (MuJoCo Menagerie)

The build on this site: Teleoperate two arms at once with ALOHA

OpenMANIPULATOR-X

• open_manipulator (ROS 2)

  The official ROS 2 packages. Start with open_manipulator_x_gazebo.launch.py.

  URDF (xacro), Gazebo, by ROBOTIS

• robotis_mujoco_menagerie: robotis_open_manipulator_x

  ROBOTIS's own MuJoCo model.

  MJCF, MuJoCo, by ROBOTIS

The build on this site: Run ROBOTIS OpenMANIPULATOR with ROS 2 and MoveIt

Unitree Go2, G1 and H1

• unitree_rl_gym

  Unitree's own loop: train in Isaac Gym, check the policy in MuJoCo, then run it on the robot. Covers Go2, G1, H1 and H1_2.

  URDF, MJCF, Isaac Gym, MuJoCo, by Unitree

• unitree_ros: robots

  Official robot description files, including go2_description, g1_description and h1_description.

  URDF, MJCF, Anything that loads URDF, MuJoCo, by Unitree

• MuJoCo Menagerie: unitree_go2

  Simplified Go2 model made from Unitree's URDF.

  MJCF, MuJoCo, by Google DeepMind (MuJoCo Menagerie)

• MuJoCo Menagerie: unitree_g1

  Simplified 29-DoF G1 made from Unitree's own model. A version with hands is included.

  MJCF, MuJoCo, by Google DeepMind (MuJoCo Menagerie)

• MuJoCo Menagerie: unitree_h1

  Made from the URDF Unitree provided. BSD-3-Clause.

  MJCF, MuJoCo, by Google DeepMind (MuJoCo Menagerie)

• Isaac Sim robot library: Go2

  In Isaac Sim at Unitree/Go2/go2.usd.

  USD, Isaac Sim, Isaac Lab, by NVIDIA

• Isaac Sim robot library: G1 and H1

  In Isaac Sim at Unitree/G1/g1.usd and Unitree/H1/h1.usd.

  USD, Isaac Sim, Isaac Lab, by NVIDIA

Builds on this site: Train a walking policy for Unitree Go2, H1 or G1 in sim · Write your own controller for a Unitree robot with SDK2

Reachy Mini

• Reachy Mini simulation guide

  Built into the SDK. Install reachy-mini[mujoco], run reachy-mini-daemon --sim, and apps run in the sim window. On a Mac, launch it with mjpython.

  MJCF, URDF, MuJoCo, by Pollen Robotics

Builds on this site: Set up Reachy Mini as a talking desk companion · Give Reachy Mini a real-time voice and personality · Make Reachy Mini move with emotion in real time · Turn Reachy Mini into a fully local baby companion

Stretch 3

• stretch_mujoco

  Hello Robot's sim stack, with cameras, lidar, a ROS 2 package and RoboCasa kitchens to drop the robot into.

  MJCF, MuJoCo, by Hello Robot

• MuJoCo Menagerie: hello_robot_stretch_3

  Model provided by Hello Robot. Apache 2.0.

  MJCF, MuJoCo 3.3.0+, by Google DeepMind (MuJoCo Menagerie)

The build on this site: Tell Stretch 3 to pick up objects around your home

TurtleBot 4

• turtlebot4_simulator

  The official sim. Install with sudo apt-get install gz-harmonic ros-jazzy-turtlebot4-simulator. Setup steps are in the TurtleBot 4 user manual.

  URDF (xacro), Gazebo worlds, Gazebo Harmonic, ROS 2 Jazzy, by Clearpath Robotics / TurtleBot

Setup guide: TurtleBot 4 user manual, simulator page ↗

The build on this site: Map your home and navigate with TurtleBot 4

Open Duck Mini

• Open_Duck_Mini repo: open_duck_mini_v2

  Robot files for the v2 duck.

  URDF, MJCF, MuJoCo, by Antoine Pirrone

• Open_Duck_Playground

  Where the walking policies are trained now. The main repo says it moved here.

  MJCF, MuJoCo Playground, by Antoine Pirrone

The build on this site: Build Open Duck Mini, a Disney BDX-style walking droid under $400

ToddlerBot

• toddlerbot repo: descriptions

  The project's own robot and scene files for each ToddlerBot variant.

  MJCF, MuJoCo, by Haochen Shi (Stanford)

• MuJoCo Menagerie: toddlerbot_2xc

  The 2XC variant, exported from the open OnShape CAD. There's a toddlerbot_2xm folder too.

  MJCF, MuJoCo, by Google DeepMind (MuJoCo Menagerie)

The build on this site: Build ToddlerBot, a low-cost open humanoid for learning

Berkeley Humanoid Lite

• Berkeley-Humanoid-Lite repo

  Training tasks for Isaac Lab, a sim-to-sim check, and robot files in all three formats under source/berkeley_humanoid_lite_assets.

  URDF, MJCF, USD, Isaac Lab, by Hybrid Robotics (UC Berkeley)

The build on this site: Build Berkeley Humanoid Lite, a sub-$5,000 open humanoid