Improving emergency evacuation planning with mobile phone location data

Timely responses to emergencies are critical for urban disaster and emergency management, particularly in densely populated mega-cities. Researchers and personnel involved in urban emergency management nowadays rely on computers to carry out complex evacuation planning. Agent-based modeling, which supports the representation of interactions among individuals and between individuals and their environments, has become a major approach to simulating evacuations wherein spatial–temporal dynamics and individual conditions need attention, such as congestion in urban areas. However, the development of optimal evacuation plans based upon agent-based evacuation simulations can be very time-consuming. In this study, to shorten the computation time to provide a timely response in an efficient way, we develop a knowledge database to store evacuation plans for typical population distributions generated by mobile phone location data. Subsequently, we use the prepared knowledge database (offline) to accelerate real-time (online) processes in searching for near-optimal evacuation plans. Our experimental result demonstrates that the evacuation plans generated with a knowledge database always outperform those that are generated without a knowledge database. Specifically, the knowledge database can reduce the computation time by an average of 96.76%, with an average fitness value improvement of 21.86%. This result confirms the effectiveness of our proposed approach in improving agent-based evacuation planning. With the rapid development of human sensor data collection and analysis, the estimation of a more accurate population distribution will become easier in future. Thus, we believe that the proposed approach of developing a knowledge database based on population distribution patterns will provide a more feasible alternative solution for evacuation planning in the practice of urban emergency management.