Evaluating the Usability of a Procedural Building Generation Tool for Unreal Engine 5

Sammanfattning

Background. Manually populating virtual environments with varied building structures is a time-intensive challenge in game development. Procedural Content Generation tools offer a way to automate this process, yet many existing systems suffer from steep learning curves and non-intuitive interfaces that limit their accessibility to non-technical users. Objectives. This thesis aims to develop and evaluate the usability of a rule-based procedural building generation tool for Unreal Engine 5 that assembles user-supplied modular assets into varied building configurations through parameters exposed in the engine’s native Details Panel. In addition, the study evaluates the intuitiveness of the exposed parameters and users’ perceived creative control over the generation process. Methods. A tool was implemented using UE5’s PCG Framework and Blueprints, and evaluated through a user study conducted at Blekinge Institute of Technology. Nine participants completed three tasks of increasing complexity and responded to the System Usability Scale questionnaire, a self-constructed questionnaire evaluating parameter intuitiveness consisting of 10 questions, and a self-constructed questionnaire evaluating perceived creative control consists five questions. Results. The tool received an overall SUS score of 80, placing it in the "Good" range and surpassing the accepted usability benchmark of 68. Parameter intuitiveness mean scores ranged between 3.89 and 4.67 across all ten questions on a five-pointscale, indicating that participants generally agreed that the tool’s parameters were clearly labelled and easy to navigate. The perceived creative control questionnaire received mean scores between 4.0 and 4.44 across all five questions on a five-pointscale, indicating that participants agreed that the tool supported their creative intentions. Specifically, participants felt in control of the creative outcome, considered the available options sufficient to achieve their creative goals, and believed they could produce a unique result that reflects their personal style using the tool. Conclusions. The tool can be considered usable by its target audience across allevaluated dimensions, and the initial hypothesis that the tool would achieve a SUSscore of 68 or above was confirmed.