Konferensartikel

In disaster preparedness, agent-based simulation (ABS) is an effective tool for aggregating information on evacuees affected by disasters. An agent usually imitates a household; and its actions are normally specified by decision models based on risk perception and social elements. Assigning socially connected households to the same shelters can better utilise resources. However, by pre-assigning specified regions to shelters, social connections are often omitted when developing evacuation plans at policy level. Thus, this paper presents a method to create social-aware evacuation plans. A GIS-enabled ABS is used to estimate evacuation demand and group evacuees according to their social connection data prior to assign them to the nearest shelters. The evacuation plans generated by the proposed method are compared against the travel-cost optimisation plans solved by using a linear model. The results obtained from a case study show that the social-aware evacuation plans offer a slightly better utilisation of shelter capacity; take similar time to evacuate households; yet could save nearly three hours to achieve complete evacuation. These results seem to confirm the competitiveness of social-aware evacuation plans as an option for evacuation planning at policy level.

agent-based simulation, social connection, K-Means clustering, linear programming, evacuation and shelter planning, GeoMASON