README.md

---
license: apache-2.0
task_categories:
- robotics
tags:
- tsfile
- timeseries
- time-series
- LeRobot
- robotics
- format:tsfile
pretty_name: grab_dibble
size_categories:
- 1K<n<10K
---

# grab_dibble (TsFile)

Apache TsFile version of [`ymatari/grab-dibble`](https://huggingface.co/datasets/ymatari/grab-dibble).

## Overview

A LeRobot v2.1 single-arm manipulation dataset recorded on a SO-101 follower
robot. The arm grasps a dibble and places it into a container. Each frame holds
the 6-DOF joint positions (shoulder pan/lift, elbow flex, wrist flex/roll,
gripper) for both the commanded `action` and the observed `observation.state`,
plus three camera views (top, surface, wrist) that are stored as videos in the
original dataset.

- **Episodes:** 50
- **Frames:** 7,642
- **Sampling rate:** 30 fps
- **Tasks:** 1 — "place the dibble in the container"

## Schema (TsFile structure)

All episodes share one TsFile with `episode_index` and `task_index` as TAG
columns; query a single episode with `WHERE episode_index = N`.

- **Time** (INT64, milliseconds) — `round(timestamp * 1000)`; the source
  `timestamp` column is dropped (it equals Time / 1000).
- **episode_index** (TAG) — episode identifier.
- **task_index** (TAG) — task identifier (single task here).
- **frame_index** (INT64) — per-episode frame counter.
- **sample_index** (INT64) — source `index` column, renamed.
- **action_0..action_5** (FLOAT) — commanded joint positions.
- **observation_state_0..observation_state_5** (FLOAT) — observed joint positions.

The 6 action/state dimensions map to `shoulder_pan.pos`, `shoulder_lift.pos`,
`elbow_flex.pos`, `wrist_flex.pos`, `wrist_roll.pos`, `gripper.pos`.

## Usage

Install the Apache TsFile Python SDK (`pip install tsfile`) and read a converted file:

```python
from pathlib import Path
from tsfile import TsFileReader

path = Path("data/grab_dibble.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())
```

## Source & license

- Original dataset: https://huggingface.co/datasets/ymatari/grab-dibble
- Author / publisher: ymatari
- License: apache-2.0
- Note: camera videos (top/surface/wrist) are NOT included; see the original dataset.