Disaster Response Applications using Distributed Computing across Delay-Tolerant Networks 

Edgar Marko Trono (1461021)



Disaster response teams perform many tasks during their operations including mapping the disaster area and Family Tracing and Reunification (FTR). In this study, we address the problem of how to realize digital systems that generates disaster area maps and aids the FTR processes. Such systems should function without continuous, end-to-end communication networks, such as the Internet, nor access to Cloud-based computing resources. To solve the problem, we propose systems that use a Delay-Tolerant Network composed of on-foot responders with smartphones and leverage response vehicles as data ferries for infrastructure-less communication. Then, we leverage existing computing devices in the disaster area with a load-balancing heuristic to handle the computation requirements of our systems. To generate disaster area maps, our system executes a map inference algorithm to convert GPS traces collected by responders. For FTR, our system executes the Eigenfaces face recognition algorithm to search for missing people from evacuee records. Using experiments and simulations, we evaluate the performance of our systems and show that our mapping system, with the load balancing heuristic can reduce the processing and delivery time of disaster area maps by approximately 50% and that our proposed FTR system can handle the computation load finding missing persons via face recognition within a fast time of approximately 7 seconds. The fast processing and delivery times benefit disaster response operations in which speed is critical.