What HARSHER SLAM is?
Localization and mapping that keeps working in dust, fog, rain and snow
Dust, fog, rain and snow blind cameras and LiDAR sensors and that is exactly when autonomous machines in mining, quarrying, construction and agriculture still have to know where they are. The Radar SLAM by Harsher localizes and maps with 4D imaging radar, which sees through dust, fog, rain, snow and darkness, combined with an IMU. It gives vehicles and machines a precise 6-DoF pose and a 3D map of their surroundings when conventional perception stacks fail, including where satellite positioning is unreliable. On the Harsher quarry dataset, HARSHER SLAM tracked a 280 m drive with a trajectory error of 0.51 m (ATE RMSE, 0.18 % of the distance travelled) from radar and IMU alone, with no GNSS, LiDAR or camera.
Harsher SLAM in RTMaps
HARSHER SLAM runs inside RTMaps as a native component, with no ROS and no middleware. The package (`rtmaps_harsher_slam.pck`) provides:
- harsher_slam: radar point cloud and IMU in; 6-DoF pose, 3D map and the registered live scan out, on standard RTMaps outputs. Runs in real time with less than one CPU core.
- ros2bag_player with pointcloud2_decoder, imu_decoder and compressed_image_decoder: replay ROS 2 recordings (.db3) directly in RTMaps, without a ROS installation.
- pose_to_viewer and points_colorizer: trajectory, current pose, live scan, height-coloured map and a chase camera in the RTMaps 3D viewer, plus highlighting of moving radar targets.
- pose_tum_writer and imu_packer: trajectory export for evaluation and IMU input from separate acceleration and gyro streams.
At the end of a run, the optimized trajectory and the full-resolution map (PLY) are saved for offline use.
Inputs: radar points as `x, y, z[, intensity[, doppler]]` and IMU as accelerations and
angular rates.
Outputs: pose (`x, y, z, qx, qy, qz, qw`), map (XYZ or colour-coded
XYZRGB), registered scan, status.
Platform: RTMaps 4.13 on Linux x86_64.
harsher gallery
BENEFITS AND KEY ADVANTAGES OF THE all-weather radar slam BY harsher
- Works in dust, fog, rain and snow: 4D imaging radar keeps localization and mapping running in the conditions that blind cameras and LiDAR, day and night.
- Accurate without GNSS: 0.51 m trajectory error over a 280 m quarry drive (0.18 % of distance), from radar and IMU only.
- Runs in real time: runs as an RTMaps component across in-car, simulation, and cloud; outputs AVC, HEVC, AV1.
- Native RTMaps, no ROS: One in-process component with two inputs and standard RTMaps outputs, ready for planning, control, logging and fusion.
- Easy to evaluate: Replay your own ROS 2 recordings in RTMaps with the included player and see the map and trajectory in the RTMaps 3D viewer.
- Complete results: Live pose, map and registered scan while running; optimized trajectory and full-resolution map at the end of a run.
how does it work?
ABOUT HARSHER
Harsher builds perception for the extreme: SLAM, radar localization and mapping for machines that work in harsh environments. Our systems keep localization and mapping running in dust, fog, severe weather and GPS-denied conditions, enabling continuous operation in mining (open-pit and underground) and construction automation. Harsher works with any radar platform for robust localization, and uses LiDAR for high-accuracy mapping.
For more informations, contact Harsher and Intempora
Please, fill the form below.