In the course of the Radar in Robotics one-day workshop at ICRA 2026, we organized a radar SLAM challenge. It involved a mobile robot equipped with three high-res 4D radars driving a 2.6km track in a mixed unstructured natural and structured urban environment. The best competing team reached an impressive 2.35m average position error (w.r.t. RTK-GNSS reference.)
Another objective of the RADARIZE project was to develop algorithms that would detect people using 4D radar scans. This application of radars is crucial for safe operation of autonomous vehicles in zero-visibility conditions, where traditional optic sensors fail. Example scenarios: low visibility after blasting in an underground mine, search&rescue in the case of a mine fire.
In the RADARIZE project, one of the research questions was whether radar mapping and ego-localization can achieve accuracy and robustness comparable to the lidar counterpart. The results shown here suggest, that in difficult mining scenarios, radar can surpass lidar localization thanks to the doppler velocity measurements.
DARPA Subterranean Challenge was an international robotics competition focusing on autonomy, perception, networking, mobility technologies, and mapping underground areas in unpredictable conditions. This competition consisted of four circuits: tunnel systems, urban underground, cave networks and a combination of all of these together. We competed as the CTU-CRAS-NORLAB team, consisting of Norlab (Université Laval) and CRAS (Czech Technical University), and scored 3rd place in the Urban circuit (Systems track with physical robots) and 2nd place in the Finals in the Virtual track (digital twin of the System track).
Benefiting from the access to a sub-arctic, university-owned forest in Québec, this project developed SLAM and autonomous navigation methods robust to 4m snow accumulation, changing weather conditions, challenging traction properties, extremely low temperatures and wildlife encounters. This video demonstrates our milestone system demonstration: the robot was autonomously repeating learned trajectory during several days, while weather changed from freezing temperatures to rainy weather.
This robotic deployment was the culmination of our Search & Rescue research in the NIFTi and TRADR projects. Both projects focused on human-robot teams, developing both the robotic technology and the human interfaces tailored to the end-user: fire brigades. In 2016, the TRADR project was asked by the Italian fire brigade Vigili del Fuoco to provide 3D textured models of two partially collapsed churches: San Francesco and Sant'Agostino in Amatrice, Italy, which was hit by an earthquake. We entered San Francesco with two UGVs (ground robots) and one UAV (drone). Sant'Agostino was inspected by one UAV while two other UAVs were providing a view from different angles to facilitate maneuvering them entirely out of line of sight. The fire brigade received detailed 3D models that allowed further planning of conservation.