Forecasting the flow of urban pollution with cellular automata

Abstract

Urban pollution is a growing health hazard in many urban centres across the globe. Prominent sources of pollution include diesel and gasoline vehicles, as well as manufacturing plants, power generation processes, and other industrial activity. In order to help understand and address pollution levels, a number of cities are installing sensor arrays; these installations will in future support monitoring and tracking of pollutants, and also underpin a range of possibilities for forecasting and mitigation. In this paper we describe an approach which forecasts the future flow and intensity of pollutants around an urban area, given recent historic sensor streams. The approach employs a cellular automaton, whose parameters are learned and adapted online by an evolutionary algorithm.