Large-scale agent-based modelling of street robbery using graphical processing units and reinforcement learning

Although agent-based modelling of crime has made great progress in the last decades, drawing concrete conclusions from the modelling results that can be directly applied to real-world environments has thus far remained challenging. In order to study different hypotheses of street robbery at the scale of a realistic urban agglomeration, a model has been developed that fully incorporates the mobility behaviour of a civilian population at high spatial (1 m) and temporal (1 s) resolution, co-simulated alongside a perpetrator agent population that is endowed with the ability to learn through experience how to travel across and roam within the urban landscape, resulting in a stochastic “common-knowledge” behaviour that mimics the intelligence that real perpetrators possess about their own environments. The model is tested on a scenario developed for the City of Cape Town, South Africa, that has a population of 4.3 million. Two different perpetrator reward signals, that capture how perpetrators gauge robbery opportunities both in terms of value and probability of success, are evaluated. The results show that perpetrator agents effectively optimise for the specified reward signals. The very high granularity of the outcomes from the simulated robberies can be compared spatially and temporally with real crime data, thereby providing a simulation framework that can be of use to criminologists, urban planners and policymakers. [Display omitted] •Mobility behaviour of millions of civilian and perpetrator agents are modelled.•Large-scale case study developed for real urban area in Cape Town, South Africa.•Perpetrator agents learn how to travel across and roam within the urban landscape.•Specified reward signal dictates perpetrator behaviour; Different hypotheses tested.•Granular simulation can be compared with actual robberies spatially and temporally.