Resilient AI housing: Toward AI-assisted computational design of resilient modular housing: A hybrid conceptual and experimental framework for extreme environments

This paper presents a hybrid design methodology that fuses rule-based artificial intelligence, environmental simulation, and parametric design to design emergency shelters resistant to low-resource extreme environment contexts. Taking the Atacama Desert as a terrestrial analog of Martian and lunar environments, the investigation explores the role of performance-driven design grounded in symbolic AI logic for adaptive architectural configurations, spatially efficient planning, and passive climate response. It combines modular design grammar with environmental simulation tools to assess design alternatives based on daylight autonomy and solar gain as key performance criteria. Although we experimented with only a single extreme environment and with only a few performance measurements, the framework is to be extended to further usage and flexible modular strategies. We describe how a simple rule-based AI assistant is incorporated to refine the spatial arrangements of variant images by using performance limits to filter and rank them. Rather than automating the design, the model proposed strives for a co-creative process, with AI acting as a human-interpretable, transparent, repeatable, and easy-to-learn design guide. The result also indicates that the AI system selects building spatial patterns that are more effective in reducing heat gain and solar exposure. This paper introduces a reproducible, open-source approach to AI-enabled computation-in-design. Its long-term possibilities include adaptive architecture on Earth and in space, providing a bridge between the current generation of climate-interactive design and a future generation of space habitation systems.