THULab/svla_so101_sim_task4_v3_multiple_1

svla_so101-sim_task4_v3_multiple_1 (TsFile)

Apache TsFile version of jadechoghari/svla_so101-sim_task4_v3_multiple_1, a LeRobot v2.1 simulation dataset recorded with a simulated SO-101 arm.

Overview

This dataset is a LeRobot v2.1 dataset recorded in simulation with an SO-101 arm (so101-sim), for a single pick-and-place-style task, with per-frame 6-DOF joint state and action plus three camera streams (top, side, wrist).

• Robot: SO-101 simulation (so101-sim)
• Scale: 1,116 episodes, 401,760 frames, 1 task, 30 fps
• Split: train
• Cameras (not uploaded): top, side, wrist

Schema (TsFile structure)

The TsFile table is named svla_so101_sim_task4_v3_multiple_1.

Role	Columns
Time	Time, INT64 milliseconds
TAG	episode_index, task_index
FIELD	frame_index, sample_index
FIELD	observation_state_0 … observation_state_5, FLOAT
FIELD	action_0 … action_5, FLOAT

The original 6-element observation.state and action vectors are flattened into scalar FLOAT measurements. The source index column is retained as sample_index.

Conversion notes

• Time = round(timestamp * 1000) with millisecond precision; time restarts inside each episode, while episode_index and task_index identify the TsFile device.
• The original timestamp field is omitted because it is exactly represented by Time / 1000.
• Camera video streams are NOT included in this repository; they remain in the source videos tree.
• meta/ is mirrored from the source. Aside from the redundant timestamp column and the excluded videos, no source rows or numeric fields are dropped.

Read example

from tsfile import TsFileReader

with TsFileReader("data/svla_so101_sim_task4_v3_multiple_1.tsfile") as reader:
    print(reader.get_all_table_schemas()["svla_so101_sim_task4_v3_multiple_1"])

Source & license

• Original dataset: https://huggingface.co/datasets/jadechoghari/svla_so101-sim_task4_v3_multiple_1
• Author: jadechoghari
• License: not declared by the original dataset; please defer to the original.