Adriano Melo

Analyzing police patrol routes by simulating the physical reorganization of agents

In this article we describe a tool for assisting the investigation of different strategies for the physical reorganization of agents. We show how the tool was used in the public safety domain to help in the study of strategies of preventive policing. A society of agents that simulates criminal and police behavior in a geographical region was constructed. In this society, artificial agents representing the police are responsible for preventing crimes. The organizational structure of the police is characterized by the existence of a centralized command that has the task of distributing and redistributing the police force in a region according to an analysis on crime and the factors that influence it. The simulation of different strategies of physical reorganization is a first step to better understand the influence that different police patrol routes have on the reduction of crime rates.

A bio-inspired crime simulation model

In this paper we describe a multiagent crime simulation model that resorts to concepts of self-organizing bio-inspired systems, in particular, of the Ant Colony Optimization algorithm. As the matching between simulated and real crime data distributions depends upon the tuning of some control parameters of the simulation model (in particular, of the initial places where criminals start out), we have modeled the calibration of the simulation as an optimization problem. The solution for the allocation of criminals into gateways is also undertaken by a bio-inspired method, namely, a customized Genetic Algorithm. We show that this approach allows for the automatic discovery of gateway configurations that, when employed in the simulation, produce crime distributions that are statistically close to those observed in real data.

GAPatrol: an evolutionary multiagent approach for the automatic definition of hotspots and patrol routes

In this work, we present a novel evolutionary multiagent-based simulation tool, named as GAPatrol. Such system is devoted to the specification of effective police patrol route strategies for coping with criminal activities happening in a given artificial urban environment, which, in turn, mimics a real demographic region of interest. The approach underlying GAPatrol allows for the automatic uncovering of hotspots and routes of surveillance, which, in real life, are usually discovered by hand with the help of statistical and/or specialized mapping techniques. The qualitative/quantitative results achieved by GAPatrol in two scenarios of study are discussed here, evidencing the potentialities of the novel approach as a promising decision-support tool for police patrolling.

A crime simulation model based on social networks and swarm intelligence

Experience in the domain of criminology has shown that spatial data distribution of crime in urban centers follows a Zipf law in which few places concentrate most of the crimes while several other places have few crimes. In order to reproduce and better understand the nuances of such a crime distribution profile, we introduce in this paper a novel multi-agent-based crime simulation model that is directly inspired by the swarm intelligence paradigm. In this model, criminals are regarded as distributed entities endowed with the capability to pursue self-organizing behavior by considering their individual (local) activities as well as the influence of other criminals. Through controlled experiments with the simulation model, we could indeed observe that self-organization phenomena (i.e. criminal behavior toward crime) emerge as the result of both individual and social learning factors. At the same time, our experiments reveal that the spatial distribution of crime achieved by experimenting with the simulation model closely follows the real crime data distribution as expected.

Self-organized and social models of criminal activity in urban environments

Multi-Agent Systems (MAS) are extensively used as a tool for simulation of dynamic systems. Geosimulation is an urban phenomena approach that uses the multi-agent methodology to simulate discrete, dynamic, and event-oriented systems. Our focus in this paper is to use self-organization, specially strategies inspired by solutions from Swarm Intelligence, as well as the idea of social networks, and demonstrate their effect on learning in geosimulation agents.