Population Generation for Large-Scale Simulation

ABSTRACT Large-scale multi-agent simulations must generate agent populations that satisfy a large set of constraints in order to recreate the population structure and demographic properties of the world being simulated. The Hats Simulator is a lightweight model of an urban space in which agents engage in individual and collective activities. In Hats, we generate hundreds of thousands of agents, each of whom belongs to a subset of thousands of benign or terrorist organizations. The population of Hats and organizations must satisfy multiple local and global constraints. Populations need to be generated quickly and reliably to facilitate repeated runs of the simulation. We present two algorithms for solving this problem: one exact but slow and another much faster one based on random,bipartite graphs. Keywords: Population Generation, Simulation, Bipartite Graph