README.md

---
language:
- en
license: cc-by-4.0
multilinguality:
- monolingual
size_categories:
- 1K<n<10K
source_datasets:
- original
task_categories:
- robotics
pretty_name: MiniVLA-Nav v1
tags:
- robotics
- navigation
- imitation-learning
- vision-language-action
- isaac-sim
- nova-carter
- differential-drive
- language-conditioned
- behavior-cloning
- simulation
- object-approach
- depth
- segmentation
configs:
- config_name: default
  data_files: metadata.parquet
---

# MiniVLA-Nav v1

**A Multi-Scene Simulation Dataset for Language-Conditioned Robot Navigation**

[Paper](https://huggingface.co/papers/2605.00397)

<!-- > Ali Al-Bustami -->

---

## Demo

<video controls preload="metadata" width="100%">
  <source src="https://huggingface.co/datasets/alibustami/miniVLA-Nav/resolve/main/assets/montage_all_scenes.mp4" type="video/mp4">
  <a href="https://huggingface.co/datasets/alibustami/miniVLA-Nav/blob/main/assets/montage_all_scenes.mp4">All-scenes montage</a>
</video>

*Nova Carter navigating to named objects across all four Isaac Sim environments.*

---

## Dataset Summary

MiniVLA-Nav v1 is a simulation dataset for the **Language-Conditioned Object Approach (LCOA)** task: given a short natural-language instruction, an NVIDIA Nova Carter differential-drive robot must navigate to the named object and stop within 1 m. Data were collected inside four photorealistic NVIDIA Isaac Sim 5.1 environments (Office, Hospital, Full Warehouse, Warehouse with Multiple Shelves).

Each of the **1,174 episodes** pairs a language instruction with per-timestep, synchronized multimodal observations:

| Modality | Resolution / Shape | Format |
|---|---|---|
| Front RGB | 640 × 640 × 3, uint8 | PNG |
| Metric depth | 640 × 640, float32 (metres) | NumPy |
| Instance segmentation | 640 × 640, uint16 | PNG |
| Continuous actions (v, ω) | T × 2, float32 | NumPy |
| Tokenized actions (7×7) | T × 2, int16 | NumPy |
| Robot poses (x,y,z,qw,qx,qy,qz) | T × 7, float32 | NumPy |

All sensors operate at **60 Hz** (physics Δt = 1/60 s).

---

## Supported Tasks

- **Language-Conditioned Object Approach (LCOA)** — given a natural-language goal and front RGB-D observations, predict continuous (v, ω) or discrete 7×7 action tokens to drive a differential-drive robot within 1 m of the named object.
- **Behaviour Cloning / Imitation Learning** — dense per-step expert labels enable direct supervised training.
- **OOD Generalisation** — structured evaluation splits test template-paraphrase and object-category out-of-distribution robustness.

---

## Multimodal Observations

Each timestep provides synchronized RGB, metric depth (float32, metres), and instance segmentation. The composites below show RGB (left) and depth colormap (right) from a mid-episode step.

| Office | Hospital |
|:---:|:---:|
| ![RGB+D office](assets/rgbd_office.png) | ![RGB+D hospital](assets/rgbd_hospital.png) |

| Full Warehouse | Warehouse (Multi-Shelf) |
|:---:|:---:|
| ![RGB+D full warehouse](assets/rgbd_full_warehouse.png) | ![RGB+D warehouse shelves](assets/rgbd_warehouse_shelves.png) |

**Depth strip** — consecutive frames from an office episode, showing depth (metres) as the robot approaches the target:

![Depth strip office](assets/depth_strip_office.png)

---

## Scenes

Four photorealistic Isaac Sim environments, each with curated seen/held-out object categories:

### Office
![Contact sheet — Office](assets/contact_office.png)

### Hospital
![Contact sheet — Hospital](assets/contact_hospital.png)

### Full Warehouse
![Contact sheet — Full Warehouse](assets/contact_full_warehouse.png)

### Warehouse (Multiple Shelves)
![Contact sheet — Warehouse Multi-Shelf](assets/contact_warehouse_multiple_shelves.png)

| Scene | Episodes | Seen Categories | Held-out Categories |
|---|---|---|---|
| Office | 281 | chair, sofa, table, monitor, plant, trash\_can | fire\_extinguisher, whiteboard |
| Hospital | 22 | chair, trash\_can | fire\_extinguisher, whiteboard |
| Full Warehouse | 54 | shelf, rack | barrel |
| Warehouse (Multi-Shelf) | 68 | shelf, rack | barrel |

---

## Object Categories

12 categories total — 9 seen during training, 3 held out for OOD evaluation.

**Seen categories:**

| chair | monitor | table | trash can |
|:---:|:---:|:---:|:---:|
| ![chair](assets/sample_chair.png) | ![monitor](assets/sample_monitor.png) | ![table](assets/sample_table.png) | ![trash can](assets/sample_trash_can.png) |

| rack | crate | shelf | barrel (OOD) |
|:---:|:---:|:---:|:---:|
| ![rack](assets/sample_rack.png) | ![crate](assets/sample_crate.png) | ![shelf](assets/sample_shelf.png) | ![barrel](assets/sample_barrel.png) |

**Held-out (OOD):** fire\_extinguisher, whiteboard, barrel — appear only in `test_ood_obj` split.

---

## Object Category Demo

<video controls preload="metadata" width="100%">
  <source src="https://huggingface.co/datasets/alibustami/miniVLA-Nav/resolve/main/assets/montage_office_categories.mp4" type="video/mp4">
  <a href="https://huggingface.co/datasets/alibustami/miniVLA-Nav/blob/main/assets/montage_office_categories.mp4">Office categories montage</a>
</video>

*All object categories navigated to in the Office scene.*

---

## Dataset Structure

```
v1/
├── dataset_meta.json            # Global metadata (scenes, camera, action space, splits)
├── assets/                      # README visual assets
├── splits/
│   ├── train_id.txt             # 261 episode IDs
│   ├── val_id.txt               #  41 episode IDs
│   ├── test_id.txt              #  50 episode IDs
│   ├── test_ood_obj.txt         #  37 episode IDs  (held-out object categories)
│   └── test_ood_lang.txt        #  36 episode IDs  (paraphrase OOD templates)
├── targets_office.yaml          # Per-scene object catalogs (3-D centroids)
├── targets_hospital.yaml
├── targets_full_warehouse.yaml
├── targets_warehouse_multiple_shelves.yaml
└── episodes/
    └── ep_{N:06d}/
        ├── meta.json                 # Full episode metadata
        ├── rgb_front/{t}.png         # 640×640 RGB frame at step t
        ├── depth_front/{t}.npy       # 640×640 float32 depth (m) at step t
        ├── seg_front/{t}.png         # 640×640 uint16 instance segmentation at step t
        ├── actions_continuous.npy    # (T, 2) float32 — (v_t, ω_t)
        ├── actions_tokens.npy        # (T, 2) int16  — discretized 7×7 tokens
        └── poses.npy                 # (T, 7) float32 — (x,y,z,qw,qx,qy,qz)
```

### Episode Metadata (`meta.json`)

Each episode's sidecar JSON records the full configuration:

```json
{
  "episode_id": "ep_000321",
  "scene_id": "full_warehouse.usd",
  "goal": {
    "target_category": "crate",
    "target_id": "crate_038",
    "goal_position_xyz_m": [-15.08, 10.77, 2.93]
  },
  "instruction": {
    "text": "Go to the crate.",
    "template_id": "train_01"
  },
  "spawn": { "tier": "mid", "spawn_to_target_dist_m": 3.574 },
  "rollout": {
    "num_steps": 219,
    "terminated_by": "success",
    "success": true,
    "collision_count": 0,
    "final_ne_m": 0.966,
    "trajectory_length_m": 2.61
  }
}
```

---

## Splits

| Split | Episodes | Description |
|---|---|---|
| `train_id` | 261 | Seen objects, seen instruction templates |
| `val_id` | 41 | Seen objects, seen templates (validation) |
| `test_id` | 50 | Seen objects, seen templates (held-out test) |
| `test_ood_obj` | 37 | **Held-out object categories** (fire extinguisher, whiteboard, barrel) |
| `test_ood_lang` | 36 | **Paraphrase OOD** instruction templates |
| **Total** | **425** | (current snapshot; full budget: 2,000) |

---

## Language Instructions

Instructions are generated from slot-fill templates with `{object}` and `{color}` placeholders.

**18 training templates** (T1–T18), examples:
- "Go to the {object}."
- "Drive to the {object} and stop."
- "Approach the {object}."
- "Navigate to the {object}."
- "Your destination is the {object}."

**12 paraphrase-OOD templates** (O1–O12), examples:
- "Make your way to the {object}."
- "Proceed to the {object}."
- "Find the {object} and come to a stop."
- "Close in on the {object}."

> **Note:** Color-slot templates are suppressed in v1 — all targets carry `color=unknown` because USD assets do not expose material-color attributes through a standard prim API. Active pool: 13 train + 10 paraphrase-OOD templates.

---

## Task Definition

**LCOA formulation:** Given instruction $\ell$ and observations $o_t = (I_t^\text{RGB}, D_t)$, output actions $a_t = (v_t, \omega_t)$ such that the robot stops within $r_\text{success} = 1.0$ m of the target object centroid.

**Action space:**
- Continuous: $(v, \omega) \in [0, 1]$ m/s × $[-1.5, 1.5]$ rad/s
- Tokenized: each dimension quantized to 7 uniform bins → 49-token vocabulary

**Episode termination:**
- **Success** — within 1 m and stationary for ≥ 5 consecutive steps
- **Collision** — stall detected (no forward progress for ≥ 16 steps near obstacle)
- **Timeout** — 1,000 steps reached without success

Only successful episodes are retained in the dataset.

---

## Spawn Tiers

Trajectory diversity is ensured through three distance tiers:

| Tier | Weight | Radius |
|---|---|---|
| Near | 30% | 1.5–3.5 m from target |
| Mid | 40% | 3.5–7.0 m from target |
| Far | 30% | Global curated floor points |

Pearson correlation between spawn distance and trajectory length: **r = 0.94**.

---

## Expert Controller

The data-collection expert is a proportional controller using pixel-level target visibility from the instance segmentation mask:

- **Target visible (≥ 32 px):** angular correction from mask centroid column + depth-based speed
- **Target not visible:** bearing-only proportional law from known goal position
- **Obstacle avoidance:** speed clamped when depth in central foreground crop < 0.25 m

---

## Rollout Statistics

| Split | N | Mean NE (m) | Mean TL (m) | Mean Steps |
|---|---|---|---|---|
| train\_id | 261 | 0.967 | 2.75 | 197.6 |
| val\_id | 41 | 0.967 | 2.83 | 205.6 |
| test\_id | 50 | 0.966 | 2.74 | 190.6 |
| test\_ood\_obj | 37 | 0.967 | 2.38 | 174.7 |
| test\_ood\_lang | 36 | 0.967 | 3.07 | 229.7 |

NE = final navigation error (distance to goal at termination). TL = trajectory length.

---

## Collection Setup

| Property | Value |
|---|---|
| Simulator | NVIDIA Isaac Sim 5.1.0-rc.19 |
| Robot | NVIDIA Nova Carter (differential-drive) |
| Camera | front\_hawk/right stereo camera |
| Physics rate | 60 Hz (Δt = 1/60 s) |
| Image resolution | 640 × 640 px |
| Random seed | 42 |
| Generation date | 2026-04-22 |

---

## Loading the Dataset

```python
import json
import numpy as np
from pathlib import Path
from PIL import Image

root = Path("v1")

# Load split
with open(root / "splits" / "train_id.txt") as f:
    train_ids = [line.strip() for line in f]

# Load an episode
ep_dir = root / "episodes" / train_ids[0]
meta = json.loads((ep_dir / "meta.json").read_text())

instruction = meta["instruction"]["text"]             # "Go to the monitor."
actions = np.load(ep_dir / "actions_continuous.npy")  # (T, 2) float32
tokens  = np.load(ep_dir / "actions_tokens.npy")      # (T, 2) int16
poses   = np.load(ep_dir / "poses.npy")               # (T, 7) float32

# Load frame t=0
rgb   = np.array(Image.open(ep_dir / "rgb_front" / "0.png"))   # (640, 640, 3)
depth = np.load(ep_dir / "depth_front" / "0.npy")              # (640, 640) metres
seg   = np.array(Image.open(ep_dir / "seg_front" / "0.png"))   # (640, 640) instance IDs
```

---

## Citation

If you use MiniVLA-Nav v1 in your research, please cite:

```bibtex
@article{albustami2026minivlanav,
  title   = {{MiniVLA-Nav v1}: A Multi-Scene Simulation Dataset for
             Language-Conditioned Robot Navigation},
  authors  = {Ali Al-Bustami and Jaerock Kwon},
  year    = {2026},
  url     = {https://huggingface.co/papers/2605.00397},
  note    = {Thesis project, Department of Robotics Engineering}
}
```

---

## License

This dataset is released under the [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/) license.

---

## Contact

Ali Al-Bustami - abustami@umich.edu