THULab/so101_single_arrange_yrpblock_top2bottom

SO101 Single Arrange YRP Block Top2Bottom (TsFile)

Apache TsFile version of skkuprism/SO101-single-arrange_YRPblock_top2bottom.

The source dataset is a LeRobot v3.0 SO-101 follower robot dataset. Its original dataset card does not declare a license. This repository stores the numeric time-series features as Apache TsFile and keeps visual streams in the original Hugging Face dataset.

Modalities: Time-series. The converted table includes robot state/action, end-effector pose, gripper state, skill metadata, task index, frame metadata, and episode tags.

Source Dataset

• Original dataset: skkuprism/SO101-single-arrange_YRPblock_top2bottom
• Codebase version: LeRobot v3.0
• Robot type: so101_follower
• Split: train (0:30)
• Episodes: 30
• Frames: 25,478
• Sampling rate: 30 fps
• Tasks: 1
• Source frame shards: 30 Parquet files
• Source data path: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• Source video path: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

Converted Files

• TsFile: data/so101_single_arrange_yrpblock_top2bottom.tsfile
• Rows: 25,478
• Table name: so101_single_arrange_yrpblock_top2bottom
• Time precision: milliseconds
• Mirrored metadata: meta/, with meta/info.json updated for the TsFile artifact

Schema

Time is generated as round(timestamp * 1000) milliseconds. Time restarts within each episode, and devices are identified by the original LeRobot tag columns.

• TAG columns: episode_index, task_index
• FIELD metadata columns: frame_index, sample_index
• FIELD scalar skill columns: skill_natural_language, skill_verification_question, skill_type, skill_progress
• FIELD scalar robot columns: observation_gripper_binary, skill_goal_position_gripper
• FIELD vectors: observation_state_0..5, action_0..5, observation_ee_pos_robot_xyzrpy_0..5, observation_radian_state_0..5, observation_radian_action_0..5, observation_radian_state_urdf0_0..5, observation_radian_action_urdf0_0..5, skill_goal_position_joint_0..5, skill_goal_position_world_xyzrpy_0..5, and skill_goal_position_robot_xyzrpy_0..5

The 6-element joint vectors use the source SO-101 joint order: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, and gripper.pos.

Conversion Notes

• The source timestamp column is dropped after being mapped to Time; it is recoverable as Time / 1000 seconds.
• The source index column is renamed to sample_index.
• Vector/list columns are flattened by preserving the full source column name, replacing . with _, and appending the element index.
• Source video features are not uploaded here: observation.images.realsense and observation.images.innomaker. Use the original dataset's videos/ tree for the visual streams.
• Aside from the redundant timestamp column and omitted video files, no numeric time-series rows are intentionally dropped.

Read Example

from tsfile import TsFileReader

path = "data/so101_single_arrange_yrpblock_top2bottom.tsfile"

with TsFileReader(path) as reader:
    tables = reader.get_all_table_schemas()
    print(tables.keys())