Vehicular Cloud Computing for Population Evacuation Optimization

This study focuses on online evacuation management. We design a framework to solve dynamic population evacuation (DPE) problems by creating an initial evacuation plan and employing inter-vehicle communication under a cloud computing architecture. For creating an initial plan, we formulate and solve dynamically (i) the shelter allocation problem (SAP) to choose a shelter node representing safety, and (ii) dynamic traffic assignment (DTA) to choose the best path leading towards the chosen destinations. The DPE process then begins with the planning phase to design a plan for the evacuation operation. Afterward, a vehicular ad hoc network (VANET) is deployed according to vehicular cloud computing (VCC) to make it possible for evacuees to share information and have online guidance from the server. We use vehicular communication to revise vehicle decisions initially taken by the plan, considering the dynamic progression of the catastrophe risk and congestion levels. To approve our new framework, we apply it to a real test case of Luxembourg city. We measure the clearance time and the mean evacuation time of the proposed solution and compare them with the results of other planning methods in the literature. Results show that online solving of DPE can minimize mean evacuation time and network clearance time by more than 10% compared to other models with only planning optimization. In addition, we analyze the performance of the proposed framework by changing the penetration rate of connected vehicles, and we observe that the increase in penetration rate leads to a decrease in network clearance time.