Agent Learning and Adaptation in a RA/CA Crime Simulation Model

Xuguang Wang, University of Cincinnati
Lin Liu, University of Cincinnati
John E. Eck, University of Cincinnati

ABSTRACT
Concern about crime displacement calls for better understanding of spatial learning and adaptation behaviors of criminals and victims. Recent advancements in the fields of spatial cognition and artificial intelligence allow us to build simulation models based on "intelligent" agents with capabilities in navigation, way finding, goal pursuing, learning and adaptation from past experiences. This paper examines the application of a multi-agent based model to simulate crime events. There are three types of agents in this model: criminals, targets and law enforcement agents. All agents are able to learn and adapt. The model has two main components. One simulates routine activities of the different agents; the other generates stochastic crime events as a results of the interactions among these agents during their daily routine activities. Through experimenting with rules and policies of individual agents' adaptation and examining the resulting offending and victimization dynamics, we have a tool that has potential of helping us evaluate the effectiveness of crime prevention policies, as well as victim self-protection policies.

(Return to Program Resources)

Updated 05/20/2006