Enhancing housing density solutions through algorithmic iterations and computational tools

Abstract

This dissertation explores the application of computational design tools to address three critical challenges impacting the future of housing development in Hawai‘i: affordability, availability, and identity. As housing costs continue to rise and pressures on land use intensify, traditional design and development processes have proven too slow and inflexible to meet growing community needs. In response, this research proposes a generative, rule-based computational model capable of producing rapid, zoning-compliant accessory dwelling unit (ADU) designs customized to specific site conditions, regulatory frameworks, and cultural contexts.The Parametric Housing Aggregation Model (PHAM) is organized into four primary clusters—Lot Definition, Module Definition, Aggregation, and Exterior Definition—each responsible for translating zoning rules, programmatic requirements, and spatial configurations into parametric operations. Utilizing Rhinoceros 3D, Grasshopper, and the Wasp plug-in, the model allows users to generate hundreds of ADU iterations in real time. It automatically tests and prioritizes designs based on surface area-driven material cost estimates, enabling users to identify low-cost configurations quickly. These cost analyses apply local construction cost data sourced from Hawai‘i-specific markets, further grounding the model's outputs in regional realities. Beyond cost and spatial efficiency, the model integrates AI-assisted visualization tools to address the challenge of architectural identity. AI renderers, including PromeAI and Midjourney, are used to generate stylistic renderings from parametric massings, helping designers and communities visualize how proposed densification strategies can maintain or adapt local architectural character. In doing so, the model not only accelerates technical design workflows but also facilitates a critical dialogue between form, affordability, density, and culture. Application of the model across test sites demonstrates its ability to perform rapid cost comparisons, simulate increased residential density under speculative policy changes, and visualize how ADUs might integrate into Hawai‘i's diverse urban and suburban contexts. Although the system is intended as a design exploration and decision-support tool—not a replacement for detailed architectural or engineering documentation—it significantly streamlines early-stage processes and enhances transparency in evaluating design trade-offs. Ultimately, this dissertation contributes to ongoing efforts to rethink housing production in Hawai‘i by offering a scalable, adaptable, and locally responsive computational framework. It demonstrates that computational tools can serve not just as generators of form, but as platforms for advancing affordability, availability, and cultural identity in the built environment.