Disaster Autonomous Aerial Response Technology

DAART (Disaster Autonomous Aerial Response Technology) is an advanced UAV autonomy framework for disaster response missions. It integrates state-of-the-art AI techniques to enable highly autonomous, resilient, and transparent multi-UAV operations in complex, dynamic environments. DAART combines adaptive multi-modal sensor fusion, deep learning-based semantic segmentation, hierarchical anomaly detection, decentralized swarm coordination, case-based reasoning for high-level decision making, reinforcement learning for navigation, and explainable AI for human-machine teaming. The sensor fusion module adaptively combines visual, thermal, LiDAR, and other sensor data to maintain robust situational awareness in degraded conditions. Semantic segmentation identifies mission-relevant objects while anomaly detection ensures reliable operation. Decentralized coordination enables scalable, fault-tolerant collaboration of UAV swarms. Sim-to-real transfer learning accelerates policy development for navigation in novel environments. Intuitive human-machine interfaces and explainable AI foster trust and effective intervention. DAART achieves unprecedented autonomy and resilience for UAVs in disaster response while meeting strict SWaP constraints. Phase I funding will support prototyping and feasibility demonstration in high-fidelity simulation. Our UAV autonomy architecture, SCOUT, which has been showcased to NASA stakeholders through demonstrations on operational UAVs in flight, and our aircraft Probabilistic Roadmap Path Planner, which plans aggressive, collision-free 3D routes in complex environments in close proximity with moving objects (useful in disaster management, public safety, and emergency response), demonstrate our proficiency in developing advanced UAV coordination systems, efficient and reliable perception and navigation solutions, robust anomaly detection algorithms, and intuitive human-machine interaction interfaces.