Procedural Building Generation from GIS Data in Unreal Engine 5

This thesis presents a methodology for creating and editing realistic buildings as part of a broader plan to develop a 3D production pipeline for railroad simulations at Corys. The goal is to develop a tool integrated in Unreal Engine 5, the software used by the company, ensuring the creation of highly detailed environments while prioritizing real-time performance and efficient editing workflows. To achieve this, a comparative analysis of existing procedural generation tools was conducted, identifying their strengths and limitations. This served as the foundation for designing a solution tailored to the needs of the company. Building on a previous internship focused on procedural modeling of a single building, this thesis extends the system to support the generation of multiple buildings using real-world geographical data. Buildings are generated using the following attributes: footprint, façade height, façade material, ground and upper floor types, roof type, roof height, and roof material. Custom rules and 3D assets were also developed to maintain architectural coherence and variety among buildings. Additionally, new editing modules were introduced to extend the existing toolset. Results show a significant improvement in the edition options and realism, in both daytime and nighttime settings, compared to the previous methodology. This improvement is achieved with minimal computational overhead, even when managing large numbers of instances. This research could be extended to support the generation of cities with consistent patterns even when input data is incomplete, along with further improvements in realism, editing capabilities, and integration with Unreal Engine’s Procedural Content Generation Framework.