# Husky UAL LiDAR-Inertial Dataset, sequences 2026-11-02: map building + TPS-A* autonomous navigation

Two sequences of a Clearpath Husky A200 in the park next to the CIDU building of the University
of Almería campus (Spain), recorded on 2026-11-02 with ROS 2 Jazzy, with all the robot sensors
(two 3D LiDARs, stereo camera, IMUs, wheel odometry, RTK GNSS):

1. `husky_2026-11-02_17-50-13_cidu_park_mapping`: teleoperated drive (5 min, ~240 m) used to
   build the point-cloud map of the area.
2. `husky_2026-11-02_18-05-02_cidu_park_tps_astar_nav`: autonomous navigation (13.5 min, ~295 m)
   with the TPS-A* planner, localizing with MOLA on the map built from sequence 1, among people
   walking by. This is the real-robot experiment of the TPS-A* path planning paper; it includes
   the planner, follower and localization outputs recorded on the robot.

`husky_2026-11-02_17-50-13_map_cidu_bueno.zip` is the map of the area built from sequence 1 (its
latest version, plus the exact one the robot used to localize and navigate in sequence 2).

- Dataset page (all sequences): https://w3.ual.es/~jlblanco/datasets/husky-ual-lidar-inertial-dataset/
- TPS-A* planner: https://github.com/MRPT/mrpt_path_planning (ROS 2 node `mrpt_tps_astar_planner`
  in https://github.com/mrpt-ros-pkg/mrpt_navigation)
- MOLA (mapping and localization): https://github.com/MOLAorg/mola

## License

This dataset is released under the
[Creative Commons Attribution 4.0 International license (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/).
You may share and adapt it for any purpose, including commercially, as long as you give
appropriate credit. Please cite it as:

> J.L. Blanco-Claraco, "Husky UAL LiDAR-Inertial Dataset", University of Almería, 2026.
> https://w3.ual.es/~jlblanco/datasets/husky-ual-lidar-inertial-dataset/

## Contents

| Path | Description |
|---|---|
| `husky_2026-11-02_17-50-13_cidu_park_mapping/` | ROS 2 bag (MCAP, uncompressed, 15.1 GB, 296 s) |
| `husky_2026-11-02_18-05-02_cidu_park_tps_astar_nav/` | ROS 2 bag (MCAP, uncompressed, 41.8 GB, 813 s) |
| `husky_2026-11-02_17-50-13_map_cidu_bueno.zip` | The map of the area, for localization and navigation (176 MB) |
| `SHA256SUMS` | Checksums |

The bags were recorded with other names (`husky_2026-11-02_17-50-13_map_video` and
`husky_2026-11-02_18-05-02_nav`); only the file names and `metadata.yaml` changed.

## Sensors and topics

All robot topics are under the namespace `/a200_0812`. Both sequences contain:

| Topic | Type | Rate | Sensor |
|---|---|---|---|
| `ouster/points` | `sensor_msgs/PointCloud2` | 10 Hz | Ouster OS0-32 LiDAR (also its raw `lidar_packets`, `imu_packets`, `metadata`) |
| `ouster/imu` | `sensor_msgs/Imu` | 100 Hz | IMU inside the Ouster |
| `velodyne_points` | `sensor_msgs/PointCloud2` | 10 Hz | Velodyne VLP-16 LiDAR (also its raw `velodyne_packets`) |
| `bumblebee/{left,right}/image_raw` | `sensor_msgs/Image` | 15 Hz | Bumblebee2 stereo camera (plus a 1 Hz `low_rate` copy) |
| `taobotics/imu`, `taobotics/imu/mag` | `sensor_msgs/Imu`, `MagneticField` | 150 Hz | HandsFree (Taobotics) 9-axis IMU |
| `platform/odom`, `platform/joint_states` | `nav_msgs/Odometry`, `JointState` | 18 Hz | Wheel odometry and encoders |
| `platform/odom/filtered` | `nav_msgs/Odometry` | 44 Hz | Robot EKF (wheels + IMU) |
| `gnss/f9p` | `sensor_msgs/NavSatFix` | 1 Hz | u-blox ZED-F9P, RTK corrections from an NTRIP caster (plus raw `ubx_*` messages) |
| `gnss/ublox` | `sensor_msgs/NavSatFix` | 1 Hz | Low-cost u-blox 7 receiver (plus NMEA messages) |
| `platform/cmd_vel`, `joy_teleop/*` | | | Velocity commands, joystick |
| `tf`, `tf_static`, `robot_description` | | | Transforms and robot model (sensor poses in `tf_static`) |

The navigation sequence also contains the outputs of the navigation stack running on the robot:

| Topic | Description |
|---|---|
| `lidar_odometry/*`, `state_estimation/*` | MOLA LiDAR-inertial localization on the prebuilt map (poses, geo-referenced poses, quality) and the `map -> odom` transform in `tf` |
| `nav_map/*` | The map loaded by the navigation stack (geo-reference, 2D obstacle points) |
| `mrpt_tps_astar_planner_node/*` | TPS-A* planner: committed and tentative paths, navigation status, footprint |
| `mrpt_trajectory_follower/*` | Path follower: reference path, predicted trajectory, status (e.g. "Goal reached") |
| `nav/waypoints*`, `nav/local_obstacles` | Goals given by the operator, live LiDAR obstacles used by the planner |
| `cmd_vel` | Velocity commands of the follower |
| `/costmap_0`, `/costmap_1` | Planner costmaps |
| `/rosout` | Logs of all the nodes |

Messages of non-standard packages (`ublox_ubx_msgs`, `mrpt_nav_interfaces`, `mrpt_msgs`,
`ouster_sensor_msgs`, `rtcm_msgs`, ...) can be decoded without those packages from the
schemas embedded in the MCAP file (e.g. with the `mcap` and `mcap-ros2-support` Python packages).

## Map (`husky_2026-11-02_17-50-13_map_cidu_bueno.zip`)

Built from sequence 1 with MOLA (LiDAR-inertial odometry with the Ouster, smoother with wheel
odometry, geo-referenced with the F9P fixes). The map frame is ENU, with the origin given in
`map.georef.yaml`.

| File | Description |
|---|---|
| `map.mm` | MOLA metric map (`mola::IncrementalPointCloud`) used for localization |
| `obstacles2d.mm` | 2D obstacle points used by the planner (dynamic objects removed) |
| `map.georef.yaml` | Geo-reference of the map frame (WGS84) |
| `map.simplemap`, `map_Images/` | Keyframes (poses + LiDAR scans) the map was built from |
| `trajectory_enu.tum`, `trajectory.tum`, `trajectory.kml` | Mapping trajectory (ENU, local, Google Earth) |
| `preview.png`, `preview_obstacles.png`, `nav_view_map*` | Top views of the map |
| `build_info.yaml`, `sm2mm_*.yaml` | How the map was built |
| `as_used_on_2026-11-02/` | The version the robot used in sequence 2 (see below) |

The top-level files are the latest version of the map, with the geo-reference recomputed on
2026-11-06. `as_used_on_2026-11-02/` holds the first version, which the robot loaded during the
navigation sequence: the `map` frame of its `tf`, poses and paths is that one. Both frames are ENU,
with origins about 0.3 m apart.

## Navigation experiment

The operator gave 15 goals one after another from a web view; the robot reached 8 of them
(about 210 m of autonomous driving), with people walking by and blocking the path at times.
Setup: TPS-A* with forward and reverse circular PTGs plus an in-place rotation one
(v_max 1 m/s, w_max 60 deg/s), a 0.30 m / 15 deg lattice, a 1.04 x 0.72 m rectangular
footprint, and as obstacles the 2D layer of the map plus the latest scans of both LiDARs.
Everything ran on the robot computer (Intel Core i7-8565U).

## Notes

- **Clock**: stamps are the ROS time of the robot computers. The Ouster was not yet synchronized
  by PTP or a GNSS pulse (set up in later sequences): use its per-point time field with care
  when de-skewing.
- **RTK GNSS**: the F9P was in RTK fixed only part of the time while moving; check the carrier
  solution in `gnss/f9p/ubx_nav_pvt` before using its fixes as ground truth.
- **Privacy**: the camera images show people walking in the park. If you find yourself in this
  dataset and want the data removed, please get in touch.
