Route recommendation for evacuation networks using MM PP/M/1/N queueing models

While natural disasters such as hurricanes, tsunamis, and earthquakes are inevitable, improving existing emergency response approaches and technology becomes critical in today's smart city development. That is, ensuring evacuees get a fast and effective evacuation route plays a very important role in increasing survival rate and reducing damage during emergency or disaster situations for smart cities. In this research, we investigate the design and development of efficient evacuation routes. By considering the realistic response to emergency evacuations, we present a model based on the number of incoming vehicles in different periods of time. In this paper, we propose a Queueing-based Recommendation system (Q-Recomm) for dynamic route recommendations for evacuees to use at the time of a disaster. Compared with other existing studies that use static conditions, our proposed algorithms consider dynamic road conditions by using queueing theory and graph theory. Specifically, we first characterize the evacuation network of a disaster as a directed acyclic graph and then exploit an MMPP/M/1/N queueing model to derive the optimal route in terms of the minimum driving time by analyzing road congestion, vehicle wait time, and road capacity as well as vehicle arrival rates for chosen routes, using high-end sensors for dynamic data acquisition. We propose the Queueing-based Time Constrained Route Recommendation (Q-TCRR) algorithm and the Queueing-based Minimum Driving Time Route Recommendation (Q-MDTRR) algorithm for route recommendation in evacuation networks. Our experimental results validate the accuracy of Q-TCRR and Q-MDTRR, respectively, and we also demonstrate that the Q-MDTRR algorithm has a lower driving time compared with both MacroServ and the shortest path route based on Dijkstra. Thus, while edge computing in vehicular ad hoc networks becomes a promising solution for evacuation networks, our two proposed algorithms can be flexibly realized in edge computing in vehicular ad hoc networks due to their computational efficiency.