THULab/aloha_static_screw_driver

ALOHA Static Screw Driver (TsFile)

This repository provides the numeric time-series portion of lerobot/aloha_static_screw_driver in Apache TsFile format. The original dataset was created with LeRobot for ALOHA bimanual manipulation.

Modalities: Time-series.

Source Dataset

• Original dataset: lerobot/aloha_static_screw_driver
• Robot: ALOHA
• Homepage: https://tonyzhaozh.github.io/aloha/
• Paper: Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware
• License: MIT

According to the source meta/info.json, the dataset has 50 episodes, 20,000 frames, one task, and a 50 Hz sampling rate.

Files and Scale

• data/aloha_static_screw_driver.tsfile — one TsFile containing all 20,000 numeric observations from the train split.
• meta/ — mirrored source metadata with converted TsFile metadata.

TsFile Schema

Table: aloha_static_screw_driver

Role	Columns
Time	Time (INT64, milliseconds)
TAG	episode_index, task_index
FIELD	frame_index, sample_index, next_done
FIELD	observation_state_0 through observation_state_13 (FLOAT)
FIELD	observation_effort_0 through observation_effort_13 (FLOAT)
FIELD	action_0 through action_13 (FLOAT)

Conversion Notes

• The original 14-element observation.state, observation.effort, and action vectors are flattened into scalar FLOAT measurements. Full source prefixes are retained, with . replaced by _.
• Time = round(timestamp * 1000) has millisecond precision and restarts for every episode. The redundant source timestamp field is omitted because it is recoverable as Time / 1000.
• frame_index is preserved; source index is renamed to sample_index; the boolean next.done is retained as the integer field next_done.
• The four source camera streams are not included in this repository. Videos remain available from the original dataset.

Reading the TsFile

from tsfile import TsFileReader

reader = TsFileReader("data/aloha_static_screw_driver.tsfile")
print(reader.get_all_table_schemas())

Citation

@article{Zhao2023LearningFB,
  title={Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware},
  author={Tony Zhao and Vikash Kumar and Sergey Levine and Chelsea Finn},
  journal={RSS},
  year={2023},
  url={https://arxiv.org/abs/2304.13705}
}