amiryanj/so101-sim-pick

This dataset was created using LeRobot.

Dataset Description

v1 — 82 successful pick-and-place episodes, generated in MuJoCo simulation.

An SO-101 follower arm picks a 40 mm orange ball off a table and places it to one side. There is no human teleoperation and no real hardware: every episode is produced by a scripted policy, so the dataset is cheap to regenerate and to grow.

• Homepage: https://github.com/amiryanj/robot_bazoo
• Simulator: MuJoCo 3.8.1, scene SO-ARM100/Simulation/SO101/scene_sim.xml
• License: apache-2.0

How the episodes were made

A deterministic state machine, not a neural policy and not a human:

1. the ball pose is read from the simulator (no perception in the loop),
2. damped-least-squares IK plans home → high → standoff → grasp,
3. the last leg runs as a straight Cartesian line — a joint-space swing knocks the ball 20–40 mm, per-step IK nudges it 0.2 mm,
4. the jaws close, the arm lifts 10 cm, carries, and releases.

Each episode starts at the settle at HOME, so the whole reach is in the data, not just the grasp.

One honest caveat: the grasp uses a weld

The SO-101 jaws form a shallow V. Measured over the contact set, the best opposing-normal dot product is −0.49, never near −1, so the fingers squeeze a sphere sideways and it rolls free. Rigid-body contact alone cannot hold this ball. A MuJoCo weld equality is therefore switched on when both fingers touch and released when the jaws are commanded open.

Everything else — arm dynamics, contacts, the ball's motion before and after the grasp — is ordinary physics. But a policy trained here learns when to close, not how to maintain a real friction grip. Expect that gap when transferring to hardware.

Cameras

key	view	resolution	fps
observation.images.external	fixed, looking at the workspace	480×640	30
observation.images.wrist	on the wrist, behind the jaws	480×640	30

observation.state and action are both 6 values in degrees: the five arm joints plus the gripper.

Domain randomization

Applied per episode, to avoid a policy that memorises one scene:

• ball position on the table, and radius ±2 %
• start pose ±6° on every joint
• light position and direction, headlight brightness 0.70–1.25×
• table colour ±0.12 per channel
• external camera position ±3 cm, fovy ±3°
• wrist camera position ±4 mm, fovy ±3°
• approach angle, roll, standoff and closing angle jittered per episode

The ball's colour is deliberately almost fixed (brightness 0.90–1.10× plus a whisper of channel noise). The orange is the cue the policy is meant to key on; randomizing the hue would teach it to ignore the one feature that matters.

Baseline

ACT (52 M parameters, 100 k steps, batch 8) trained on this data reaches 17/20 picks on ball positions held out from training.

Versioning

v1 is this scripted MuJoCo data. Later versions may replace the trajectories entirely; load a specific one with:

from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("amiryanj/so101-sim-pick", revision="v1")

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "robot_type": "so101_follower",
    "total_episodes": 82,
    "total_frames": 23698,
    "total_tasks": 1,
    "chunks_size": 1000,
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 200,
    "fps": 30,
    "splits": {
        "train": "0:82"
    },
    "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
    "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
    "features": {
        "observation.state": {
            "dtype": "float32",
            "shape": [
                6
            ],
            "names": [
                "shoulder_pan",
                "shoulder_lift",
                "elbow_flex",
                "wrist_flex",
                "wrist_roll",
                "gripper"
            ]
        },
        "action": {
            "dtype": "float32",
            "shape": [
                6
            ],
            "names": [
                "shoulder_pan",
                "shoulder_lift",
                "elbow_flex",
                "wrist_flex",
                "wrist_roll",
                "gripper"
            ]
        },
        "observation.images.external": {
            "dtype": "video",
            "shape": [
                480,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 480,
                "video.width": 640,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false
            }
        },
        "observation.images.wrist": {
            "dtype": "video",
            "shape": [
                480,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "info": {
                "video.height": 480,
                "video.width": 640,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "video.fps": 30,
                "video.channels": 3,
                "has_audio": false
            }
        },
        "timestamp": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null
        },
        "frame_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "episode_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "task_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        }
    }
}

Citation

BibTeX:

[More Information Needed]