Evacuation Planning via Evolutionary Computation

According to the Life Safety Code reg , the geometry of a building, the location of exits, and the number of exits dictate the means of egress for all people occupying a building. In this paper we show how evolutionary computations in the form of Genetic Algorithms and Estimation of Distribution Algorithms are used to evolve the placement of exits in order to optimize overall evacuation time. In particular, a generational GA, a steady-state GA, and an elitist EDA are used to evolve the placement of exits for two practical design problems. The algorithms are evaluated in terms of success rate, number of function evaluations, and best fitness. For both problems, the steady-state GA outperformed the other algorithms in all evaluation categories.