Optimization of the convergence of an agent-based simulation

Abstract

The aim of this term project is to develop new strategies for the agentbased simulation tool MATSim in order to increase the convergence rate and also the maximum score at the equilibrium. A new module uses the ReRouting, TimeMutator and SelectExpBeta strategies with probabilities that depend on the current iteration number; the importance of each module and in particular of the SelectExpBeta module are also discussed.