README.md

---
license: other
task_categories:
- robotics
tags:
- robotics
- agriculture
- under-canopy
- navigation
- rosbag
- terrasentia
pretty_name: LeCropFollow
configs:
- config_name: cpp
  data_files: cpp.parquet
- config_name: crow
  data_files: crow.parquet
- config_name: lecrop
  data_files: lecrop.parquet
  
---

<h1 align="center">
  LeCropFollow
</h1>

<h3 align="center">
  Latent Space Planning for Navigation in Unstructured Crop Fields
</h3>

<p align="center">
  <strong>Felipe Tommaselli</strong><sup>1</sup> &middot;
  <strong>Francisco Affonso</strong><sup>2</sup> &middot;
  <strong>Arthur Rocha</strong><sup>1</sup> &middot;
  <strong>Gianluca Capezzuto</strong><sup>1</sup><br>
  <strong>Arun Narenthiran Sivakumar</strong><sup>2</sup> &middot;
  <strong>Girish Chowdhary</strong><sup>2</sup> &middot;
  <strong>Marcelo Becker</strong><sup>1</sup>
</p>

<p align="center">
  <sup>1</sup> University of Sao Paulo &nbsp;&nbsp;
  <sup>2</sup> University of Illinois Urbana-Champaign
</p>

<p align="center">
  <em>IEEE Robotics and Automation Letters, 2026</em>
</p>

<p align="center">
  <a href="https://arxiv.org/pdf/2606.31941">
    <img src="https://img.shields.io/badge/Paper-PDF-b31b1b?style=flat-square&logo=arxiv&logoColor=white" alt="Paper">
  </a>&nbsp;
  <a href="https://arxiv.org/abs/2606.31941">
    <img src="https://img.shields.io/badge/arXiv-2026.XXXXX-b31b1b?style=flat-square&logo=arxiv&logoColor=white" alt="arXiv">
  </a>&nbsp;
  <a href="https://felipe-tommaselli.github.io/lecropfollow/">
    <img src="https://img.shields.io/badge/Project-Page-4285F4?style=flat-square&logo=google-chrome&logoColor=white" alt="Project Page">
  </a>&nbsp;
  <a href="https://youtu.be/hV1fDjQsgOs">
    <img src="https://img.shields.io/badge/Video-YouTube-FF0000?style=flat-square&logo=youtube&logoColor=white" alt="Video">
  </a>&nbsp;
  <a href="https://huggingface.co/datasets/arthurpompeu/lecrop-data">
    <img src="https://img.shields.io/badge/Data-HuggingFace-FFD21E?style=flat-square&logo=huggingface&logoColor=white" alt="Data">
  </a>&nbsp;
  <a href="https://api.wandb.ai/links/lecropfollow/mwd63kw7">
    <img src="https://img.shields.io/badge/Models-W%26B-FFBE00?style=flat-square&logo=weightsandbiases&logoColor=white" alt="Models">
  </a>
</p>

---

## Models

Please check the `Files and versions` for our most up-to-date models. For more information, check: https://github.com/Felipe-Tommaselli/lecropfollow

## Data

Navigation datasets from a **TerraSentia** agricultural robot driving
**under the canopy**, extracted from ROS1 bags. Three sources/controllers:

| Config | Episodes | Description |
|---|---:|---|
| CropFollow++ (`cpp`)    | 143 | Crop-follow / pure-pursuit (logs `path`) |
| CROW (`crow`)   | 31  | iLQR controller (logs `crop_lines`, `goal`, `ilqr_time`) |
| LeCropFollow (`lecrop`) | 137 | MPPI/RL + vision (logs `dist_err`, `head_err`, `mppi_*`, `keypoint`) |

```python
from datasets import load_dataset
ds = load_dataset("arthurpompeu/lecrop-data", "cpp", split="train")
ep = ds[0]
# ep["rgb"]            -> per-episode video (camera)
# ep["odom_pos_x"], ep["odom_vel_x"], ep["cmd_lin_x"], ...  -> time series
```

## Structure

- **Each row = one episode**: a stretch where the robot drove through the field
  until it stopped for a while. Episodes were segmented from the (smoothed) odom
  speed: "moving" when `v > 0.05 m/s`; a new episode is cut when it stays stopped
  for `>= 5 s`. A single bag can yield several episodes; episodes shorter than
  `2 s` or with `< 20 messages` were dropped. Bags without odometry become a
  single episode (the whole bag).
- **Each column = one signal** (a ROS topic), stored as a **list** (the time
  series for that episode). Topics have different rates, so each one has its own
  time vector `*_t` (seconds, relative to the episode start) and its own length.
- **Videos** (`rgb`, `lidar_plot`, `keypoint_vis_*`) are the `Video` type
  (MP4/H.264): **one video per episode** at ~10 fps, with per-frame times in
  `*_t`. The HF viewer renders a player.
- **Removed**: rosbag-level fields (`header`, `seq`, `stamp`, `frame_id`,
  `covariance`, `layout`), heavy raw sensors (depth and LiDAR/PointCloud) and
  plumbing (`tf`, `camera_info`).

> **Note — browsable version.** This published version is downscaled for the
> dataset viewer: videos are re-encoded to **320×180** and each numeric signal is
> **sub-sampled to <= 250 samples per episode**. This keeps trends/shapes intact
> and makes the viewer fast, but it is **not full resolution**. For training,
> request the full-rate / 640×360 variant.

## Columns

Metadata (scalars): `source`, `episode`, `bag`, `duration_s`, `n_msgs`.

Signals (lists; `<g>_t` = relative time in s for group `<g>`):

| Group | Columns | Source |
|---|---|---|
| `odom_*`   | `odom_t`, `odom_pos_{x,y,z}`, `odom_quat_{x,y,z,w}`, `odom_vel_{x,y,z}`, `odom_angvel_{x,y,z}` | `/…/dlio/odom_node/odom` |
| `imu_*`    | `imu_t`, `imu_acc_{x,y,z}`, `imu_gyro_{x,y,z}`, `imu_quat_{x,y,z,w}` | `/…/imu` |
| `cmd_*`    | `cmd_t`, `cmd_lin_{x,y,z}`, `cmd_ang_{x,y,z}` | `/…/cmd_vel` |
| `motion_*` | `motion_t`, `motion_lin_{x,y,z}`, `motion_ang_{x,y,z}` | `/…/motion_command` |
| `path_*`   | `path_t`, `path_pos_{x,y,z}`, `path_quat_{x,y,z,w}` (list of lists: a polyline per step) | `/…/path` |
| `goal_*`   | `goal_t`, `goal_pos_{x,y,z}`, `goal_quat_{x,y,z,w}` | `/…/goal` (crow) |
| `crop_lines_*` | `crop_lines_t`, `crop_lines_{m1,b1,m2,b2}` (crop-row lines) | `/…/crop_lines` (crow) |
| `ilqr_*`   | `ilqr_t`, `ilqr_time` | `/…/ilqr_time` (crow) |
| `dist_err`, `head_err` | predicted lateral / heading error | `/…/*_error_predicted` (lecrop) |
| `mppi_dist`, `elite_scores`, `value_info` | MPPI/RL debug (lists of lists) | `/…/rl_debug/*` (lecrop) |
| `keypoint` | vision keypoints (list of lists) | `/…/vision/keypoint` (lecrop) |

Videos (`Video`, MP4/H.264, ~10 fps; `<g>_t` = per-frame time):

| Column | Content | Source |
|---|---|---|
| `rgb` | RGB camera video | `/…/rgb/image_rect_color/compressed` (all) |
| `lidar_plot` | LiDAR plot with crop rows | `/lidar_plot` (crow) |
| `keypoint_vis_argmax`, `keypoint_vis_heatmap` | keypoint-visualization videos | `/…/vision/keypoint_vis_*/compressed` (lecrop) |

> Source-specific columns are null when an episode does not have them (e.g. some
> `cropfollowpp_lecropfollow` bags inside `cpp` carry `lecrop` columns).

## Citation

Please, consider citing our work: 
```
@ARTICLE{tommaselli2026lecropfollow,
  author={Tommaselli, Felipe and Affonso, Francisco and Rocha, Arthur and Capezzuto, Gianluca and Sivakumar, Arun Narenthiran and Chowdhary, Girish and Becker, Marcelo},
  journal={IEEE Robotics and Automation Letters}, 
  title={LeCropFollow: Latent Space Planning for Navigation in Unstructured Crop Fields}, 
  year={2026},
  volume={},
  number={},
  pages={1-8},
  doi={10.1109/LRA.2026.3710052}
}
```

## License

Code is released under the [MIT License](LICENSE). The paper is published under CC BY.