SO-101 tabletop (MuJoCo)

A LeRobot EnvHub environment: an SO-101 arm in the turquoise of the LeRobot kits, four painted wooden 3 cm cubes (red, green, blue, yellow) and a green open box on a table in a simple room, simulated with MuJoCo. Each reset places the cubes, the box and the front camera with seeded offsets.

An Isaac Sim version with the same layout, cameras, observations and actions is lerobot/so101-tabletop-isaac: the same policies and commands run against both.


Observation	`pixels`: `front` (scene camera) and `wrist` (gripper camera), 480×640 RGB; `agent_pos`: 6 joint positions
Action	6 absolute joint position targets: `shoulder_pan`, `shoulder_lift`, `elbow_flex`, `wrist_flex`, `wrist_roll` in degrees, `gripper` 0-100 (the units of LeRobot's `so101_follower` with `use_degrees=True`)
Control rate	30 fps
Task	"Put all the cubes in the green box."
Reward	the fraction of cubes in the box at each step (0, 0.25, ... 1); `info["cubes_in_box"]` gives the count
Episode end	success (terminated) when all four cubes are in the box; truncated after 3600 steps (2 minutes), set with `episode_length`
Reset	seeded cube and box poses, front camera jitter, joint noise

With the LeRobot simulator server

From a LeRobot checkout, start the server. The image installs requirements.txt, and env.py declares the control rate and joint names in LEROBOT_SIM:

lerobot-env-server --envhub=lerobot/so101-tabletop-mujoco --docker=true

Then drive it as a robot, here with a policy:

lerobot-rollout --robot.type=sim \
  --robot.env.type=envhub --robot.env.hub_path=lerobot/so101-tabletop-mujoco --robot.env.task=so101_tabletop \
  --robot.env.server.endpoint=tcp/127.0.0.1:7448 \
  --policy.path=<your SO-101 checkpoint> --fps=30

Without --robot.env.server.endpoint, the same command runs the environment in-process (with --robot.env.trust_remote_code=true and the requirements installed next to LeRobot). lerobot-eval takes the same options as --env.*, e.g. --env.type=envhub --env.hub_path=lerobot/so101-tabletop-mujoco --env.task=so101_tabletop.

Observations reach policies as observation.images.front, observation.images.wrist and observation.state; use --rename_map if your checkpoint names its cameras differently.

In-process

from lerobot.envs.configs import HubEnvConfig

envs = HubEnvConfig(hub_path="lerobot/so101-tabletop-mujoco", trust_remote_code=True).create_envs(n_envs=1)

make_env_kwargs (--env.make_env_kwargs on the command line) override the defaults of SO101Env in env.py (episode length, image size, cameras, initial joints, reset randomization). Headless machines need MUJOCO_GL=egl.

Files

File	Content
`env.py`	the EnvHub entry point (`make_env`), the Gymnasium env and `LEROBOT_SIM`
`so101_tabletop.xml`	the scene: room, table, cubes, box, lights, front camera and the arm's mount site
`robotstudio_so101/`	MuJoCo Menagerie's SO-101 model, attached at the mount site (`env.py` recolors its printed parts)
`textures/wood.png`	a seamless wood grain, tinted per cube

Licenses

This environment is Apache-2.0, including textures/wood.png (generated procedurally for it).
robotstudio_so101/ is MuJoCo Menagerie's SO-101 model (commit 0059d4335f8156206f63a35662313385f7ad6d74), derived from TheRobotStudio's SO-ARM100, under the Apache-2.0 license in that folder.