In this project, we study crime by means of agent-based simulation, focusing particularly on the study of strategies for police patrolling. By resorting to evolutionary computation (EC) resources, our main objective is to automatically uncover effective police patrol routes for coping with certain preconceived scenarios of crime occurrences that typically arise in big urban centers like Fortaleza, a metropolis with more than two million citizens in the northeast of Brazil. That is, the idea is not to design such routes of surveillance by hand, as it is normally done, but to let them emerge as a direct result of the application of a customized genetic algorithm (GA) approach.

GAs are general-purpose search and optimization algorithms that comply with the Darwinian natural selection law and with some principles of population genetics to efficiently design (quasi-)optimal solutions to complicated problems. Such metaheuristics maintain a pool of chromosomes, which represent plausible solutions to the problem, and evolve over time through a process of competition and controlled variation. The conceptualization of GAPatrol, as we name our approach, was directly inspired by the recent increasing trend on hybridizing multiagent systems (MAS) with evolutionary algorithms in such a way as to combine their positive and complementary aspects. Such philosophy of combining evolutionary and multiagent notions into a unified methodology seems to be indeed very effective, particularly in the emergent design of complimentary police patrol routes for dealing with the nuances underlying the simulated crime scenarios under consideration. Moreover, GAPatrol can be regarded as a promising candidate tool for assisting police managers in the definition of novel public-safety, preventive policies.