CROSS

This project develops intelligent aerial–ground robotic systems for mountain rescue and disaster response. We focus on perception-driven autonomy in alpine terrain, where robots must understand their environment, navigate safely, and make time-critical decisions. The approach combines multi-sensor fusion (e.g., GNSS/IMU, proximity sensors, lidar/stereo, optical/thermal cameras, ARTVA/ARVA receivers, and optional GPR) with learning-based detection and classification to identify subjects and hazards—most critically, buried avalanche victims via radio-signal comparison and fusion with visual/thermal cues. A collaborative swarm strategy coordinates aerial and terrestrial robots to accelerate area coverage, bridge line-of-sight and terrain gaps, and increase robustness under adverse weather and complex topography.

Methodologically, we define realistic alpine scenarios and evaluate: (i) risk-aware navigation and obstacle avoidance; (ii) GPS-degraded localization via visual–inertial odometry and map priors; (iii) target identification through supervised and self-supervised models trained on real and synthetic data; and (iv) multi-robot task allocation for rapid search, confirmation, and handover to first responders. Key metrics include detection/localization time, coverage efficiency, target identification accuracy, and resilience to environmental perturbations.

Expected outcomes are a validated CROSS toolkit—algorithms, datasets, and protocols—for alpine SaR operations; demonstrated reductions in victim localization time; and interfaces that integrate seamlessly with first-responder workflows. The work aligns with terraXcube’s mission on risk and disaster management, advancing autonomous rescue robotics without overlap with virtual reality, testing and industrial applications.