Kinga Aszyk
Supervisor(s): Marek Wernikowski
West Pomeranian University of Technology
Abstract: Integrating dynamic weather systems into realistic games is a key step toward enhancing both immersion and gameplay mechanics. However, typical implementations often lack the depth of the real world, relying instead on simple, random transitions between fixed conditions. This paper proposes a novel hybrid generation system, implemented in the Unity engine, designed to better reflect actual weather conditions. The architecture integrates Perlin noise for wind direction, a neural network to determine atmospheric pressure, temperature and wind speed, and Markov chains for weather state selection. To create this model, we introduce a rule-based approach for determining categorical weather states directly from continuous meteorological data. We conducted user testing to evaluate preferences between this realistic model and traditional random systems. The results indicate that increased realism is important, however its impact on the player experience remains heavily dependent on the specific game genre.
Keywords: Video Games
Full text: Year: 2026
