SinCity 2.0: An Environment for Exploring the Effectiveness of Multi-agent Learning Techniques

In this paper we present an extensive and practical analysis of Multi-agent learning strategies using our open simulator SinCity 2.0. SinCity has been developed in NetLogo and it can be considered as an extension of the simple predator-prey pursuit problem. In our case, the predators are substituted by police cars, the prey by a thief and the chase is performed in an urban grid environment. SinCity allows to model, in a graphical friendly environment, different strategies for both the police and the thief, also implying coordination and communication among the agent set. We build this model, introducing traffic and more agent’s behaviors to have a more realistic and complex scenario. We also present the results of multiple experiments performed, comparing some classical learning strategies and their performance.