Training Cellular Automata to Simulate Urban Dynamics: A Computational Study Based on GPGPU and Swarm Intelligence

We present some results of a computational study aimed at investigating the relationship between the spatio-temporal data used in the calibration phase and the consequent predictive ability of a Cellular Automata (CA) model. Our experiments concern a CA model for the simulation of urban dynamics which is typically used for predicting spatial scenarios of land-use. Since the model depends on a large number of parameters, we calibrate the CA using Cooperative Coevolutionary Particle Swarms, which is an effective approach for large-scale optimizations. Moreover, to cope with the relevant computational cost related to the high number of CA simulations required by our study, we exploits the computing power of Graphics Processing Units.