Department of Computer Science

Abstract

Cities are the largest human artifacts and appear often in digital entertainment. Building digital models of them is extremely difficult, yet there are few tools that can assist artists in this task. We describe a method for procedurally generating city models using agent- based modeling, an established simulation technique in the social, biological, and physical sciences. Our system accepts a terrain description as input and produces a land usage map that portrays roads and residential, commercial, and industrial land use. The results con- form well to modern patterns of land use, but each generated city is unique. Our system is controlled by a set of parameters that users can alter numerically or gesturally in order to constrain and guide the results to fit a particular applications needs.