Data-driven generation of virtual city layout

Procedural Content Generation (PCG) is a growing field in computer science that is progressively gaining attention. Through the automated or semi-automated creation of digital content, creativity can be increased, new experiences can be offered, and development costs can be reduced. Furthermore, it makes the fabrication of scenes and landscapes with theoretically infinite dimensions possible. The generation of virtual environments, particularly cities and urban areas, can provide numerous advantages to designers and developers. It has seen adoption in various domains, including, but not limited to, video games, movie production, and social simulations. Most existing city generation approaches investigate the artificial generation of new layouts, making only limited use of real-life data. In this thesis, we explore the direct integration of existing city layout information into the generation process. A filter function is provided to obtain the required data in the correct format. The proposed method for city layout generation simulates the attraction and adaptation of two or more urban road networks to each other. As a result, a combined layout containing interpolated features of the selected input files is generated. The implementation was done in Unity, as we rely on the internal physics engine for certain operations. User interaction is possible during the generation process, allowing structures to be modified as desired. The performance is comparable to existing simulation- and agent-based approaches for medium-sized areas. A user study confirms that the generated layouts include recognizable characteristics of the input data.