A Markov chain location–allocation meta–model for hurricane relief planning
We present a meta–model for the pre–positioning of state staging areas (SSA), activation of points of distributions (PODs), and allocation of commodities for hurricane relief. Strength and path of the hurricane are stochastic processes modelled as discrete Markov chains. Demand is also treated as a stochastic parameter. The meta–model is two models: a location model to determine SSAs to activate and a location–allocation model to determine PODs to activate and how much commodities for each of them. To solve it, we used SMOSA, a multi–objective simulated annealing heuristic, in which the initial solution and the cooling rate were determined via design of experiments. The initial temperature is irrelevant, but temperature reduction must be gradual. The Markov chains performed as well or better than forecasts made by the US National Hurricane Center (NHC), and the heuristic performed well, compared to ILP, when solving a network of 20 SSAs and 19 PODs.
Keywords: hurricane relief, stochastic Markov chains, simulated annealing, emergency response planning, location–allocation metamodels, stochastic modelling, emergency planning, emergency management, hurricanes, staging area pre–positioning, state staging areas, SSA, points of distributions, resource allocation, location models, design of experiments, DoE, disaster relief, emergency relief