A Secure End-to-End Cloud Computing Solution for Emergency Management with UAVs.

To properly determine the severity of an emergency and provide a quick response, information has to be gathered, confirmed, and analyzed. Artificial intelligence assisted unmanned aerial vehicles (UAVs) can provide a unique perspective and review extensive information for an incident. However, their limited computing capacity and battery duration raise the challenge to compute-intensive mobile applications. In addition, because they serve civilian purposes, the security of UAV communication systems has become critical. To enhance first responders' safety and effectiveness, we develop an end-to-end UAV-assisted emergency management system with secure wireless communication links, cloud-based deep learning cognitive algorithms, and easy-to-use mobile user interface. We implement our system onto a hardware platform including Infineon Larix drone, Raspberry Pi and camera, and GPU server. We evaluate the tradeoff between detection accuracy and processing time, and show that our system achieves soft real-time end-to-end latency for data transmission, processing, and reception.