Source: https://actionable.ai/datasets/R03.html

[Skip to content](https://actionable.ai/datasets/R03.html#main)

ROBOT / MULTI-STEP / AA R03

# Long-Horizon Robot Data

Multi-step, bimanual demonstrations for learning sequences of action.

[Enquire about this dataset ↗](https://actionable.ai/?dataset=R03#contact)[Browse all datasets ↗](https://actionable.ai/#catalogue)

COLLECTION INVENTORYOn enquiry

SAMPLE RECORDINGS12

Sample access, licensing, and delivery are confirmed for each enquiry.

CAPTURE DETAILS

## Sample specifications

| Capture system | Yuanli Lingji / dos_w1

| Resolution | 640 × 480 RGB

| Frame rate | 30 fps

| Format | RGB MP4 + depth MKV + action JSONL

| Views | 3 RGB + 3 depth views

| Sample coverage | 4 long-horizon recordings and 8 manipulation recordings

| Robot identifier | dos_w1

| Views | 3 RGB views and 3 depth views in original packages

| RGB capture | 640 × 480 at 30 fps

| Action streams | Dual arms: 250 Hz; mobile base: 50 Hz

| Formats | RGB MP4, depth MKV and action JSONL

INSIDE THE DATASET

## Package and coverage

RGB video, depth video and action sequences for multi-step and bimanual manipulation. Arm actions at 250 Hz; mobile-base actions at 50 Hz.

### Annotation

Descriptive task sequences; per-step timing is not included.

### Coverage

4 long-horizon and 8 manipulation sample recordings.

REPRESENTATIVE WORK

## Sample tasks

- Retrieve a power bank from a drawer and pack it into a backpack
- Pack stationery into a pencil case and return it to a drawer
- Stack blocks
- Pour particles from a cup into a box
- Bimanual bowl handover and shelf placement

EXPLORE THE CATALOGUE

## Find the right data for your model.

[Next: Stereo Ego · 1080p ↗](https://actionable.ai/datasets/E01.html)[All eight collections ↗](https://actionable.ai/#catalogue)