Datasets:
18 episodes · 30 fps · 2 cameras · 640×480 av1
SO-101 log splitter, loading and operating (ACT policy evaluation)
Autonomous evaluation episodes of an ACT policy performing the task "Load and operate logsplitter". The robot is an SO-101 arm with a miniature log splitter mounted on it. The splitter is driven by a linear motor and acts as a seventh degree of freedom, recorded as logsplitter.vel in both the action and the state.
Two cameras (wrist and side). Teleoperated training data for the same task is in logsplitter_train_batch_01 and logsplitter_train_batch_03.
Contents
- Robot:
logsplitter_follower - Episodes: 18
- Frames: 16,617
- FPS: 30
| Feature | Type | Details |
|---|---|---|
action |
float32[7] | shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos, logsplitter.vel |
observation.state |
float32[7] | shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos, logsplitter.vel |
observation.images.wrist |
video, 640x480 @ 30 fps | |
observation.images.side |
video, 640x480 @ 30 fps |
This dataset is in LeRobot format (codebase version v3.0); the full schema is in meta/info.json.
Usage
from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("primordial-spork/eval_logsplitter_act")
- Downloads last month
- 106