The first hours after a natural hazard event (e.g., a flood or earthquake) are critical to reduce damages and efficiently ramp up disaster relief operations. Relief forces, however, are challenged by various uncertainties concerning both the operational area (e.g., damages to infrastructure, need for relief goods) and the reaction of residents requiring aid (e.g., where do residents travel to and how do they influence each other). This poses the question, how can such factors be integrated in planning or training frameworks of decision makers to facilitate understanding and support operations?
One potential option is through the use of agent-based modeling. In recent work, we considered each household individually as an agent within a simulation-optimization framework. If requesting aid, this agent travels to a distribution area in close proximity based on a probability function. Furthermore, the agent is influenced by the action of other agents in the systems, which influences the selection of the distribution point over time. For instance, if an agent travels to a distribution point, but does not receive any goods due to stockouts or delayed arrival of relief goods, the agent sends a text message or posts this event on social media. This information is seen and trusted by a certain number of other nearby agents, influencing their destination choice in the future.
Additionally, a disrupted road network is modeled where each potential point of risk is indicated with an event probability (for more details refer to previous works on road disruptions and rail distruptions). In this work, we considered three different types of vehicles: (i) regular vehicles that drive a detour in case of a disruption; (ii) vehicles with special equipment that can pass a disruption at a lower speed; and (iii) vehicles with repair equipment to fix the road disruption. Based on the vehicle type and current position, the system automatically selects the best route as shown below for a special or repair vehicle (green) and a regular vehicle (blue).
By combining agent-based modeling with disrupted road settings, various disaster settings can be investigated to derive relief strategies, which consider both potential infrastructure damages and resident behavior. Nevertheless, further work is required to improve findings. Particularly, the behavior of residents and their reaction to various events (e.g., delayed shipments, information received by word of mouth or via social media) needs to be studied closely to enable the detailed implementation of such individual human factors in agent-based decision support systems. This would further allow one to differentiate between various resident groups and to model each based on their individual values and characteristics to tailor relief strategies and respond swiftly to an event.
– Fikar, C, Hirsch, P, Nolz P (2017) Agent-based simulation optimization for dynamic disaster relief distribution. Central European Journal of Operations Research, in press. [Access: Springer SharedIt]