Georgia Institute of Technology
A Multi Objective Generative Optimization Framework for Performance-Based Urban Design
Abstract
dc:description.abstractCities worldwide face the urgent challenge of achieving carbon neutrality while adapting to the increasing impacts of climate change (Citaristi, 2022; Legg, 2021). Urban environments, where buildings account for up to 60% of total carbon emissions, must reduce energy demand, transition to clean energy sources, and improve infrastructure resilience (Ürge-Vorsatz et al., 2020). Despite widespread commitments to carbon-neutrality targets, many city-level initiatives remain constrained by fragmented implementation, escalating costs, and limited decision-support capacity, resulting in slow progress toward emission reduction goals (Bulkeley & Betsill, 2013; Creutzig et al., 2018; Seto et al., 2016). In response, this dissertation’s primary contribution is a systems-level computational framework that integrates urban energy modeling, generative design, multi-objective optimization, and machine learning to support performance-based urban design and decarbonization decision-making. The framework is applied through district-scale and cross-climate case studies to demonstrate how data-driven and generative approaches can guide low-carbon, high-performance urban transformation. The proposed framework is structured around three methodological layers. First, in Chapter Three, it integrates Urban Building Energy Modeling (UBEM) to simulate building and district-scale energy use and carbon emissions under varied design, occupancy, photovoltaic, and climate scenarios. Applied to the Toyosu district in Tokyo, the UBEM analysis evaluated hundreds of scenarios and revealed non-linear relationships between density and energy use, as well as the critical role of rooftop photovoltaic coverage and efficiency in reducing district energy demand under both current and future climate conditions. Second, in Chapter Four, it develops a generative design and multi-objective optimization workflow for exploring and comparing alternative district-scale carbon-neutral regeneration pathways, using parametric and generative techniques to model urban morphology and capture key design variables such as density, building height, floor area ratio, orientation, and land-use composition that influence energy and carbon performance. Third, in Chapter Five, the framework incorporates machine learning, specifically artificial neural networks, to enable rapid urban energy prediction when physics-based simulation becomes computationally intensive. The framework is demonstrated through multiple real-world case studies, including district-scale applications in Tokyo (Toyosu and Nihonbashi) and cross-city comparative analyses in Tokyo, New York City, Atlanta, and Riyadh. In the Nihonbashi case study, the generative design and multi-objective optimization workflow evaluated 120 alternative urban regeneration scenarios across 154 buildings, revealing distinct Pareto-optimal pathways that balance energy reduction, carbon emissions, construction cost, and outdoor thermal comfort. Iterative regeneration cycles demonstrated substantial district-scale carbon reduction potential, achieving up to 97% emission reduction and highlighting that multiple decarbonization pathways can be identified rather than a single optimal solution. In the predictive modeling layer, artificial neural networks trained on UBEM-generated datasets achieved mean absolute errors of approximately 11-12 kWh/m²·year across cities, demonstrating strong predictive accuracy and robust cross-climate transferability using a limited set of urban form and system parameters. Together, these studies demonstrate a cohesive framework that transitions from descriptive urban energy analysis to predictive modeling and prescriptive design exploration. By aligning simulation, generative optimization, and machine learning within a unified computational process, this dissertation advances the field of performance-based urban design and carbon-neutral urbanism. It provides a transferable, scalable decision-support framework that enables urban designers, planners, and policymakers to evaluate trade-offs, test scenarios, and guide decarbonization strategies under uncertainty across diverse urban and climatic contexts.
Degree
thesis:*- Name thesis:degree_name
- Architecture, PhD
- Department thesis:degree_department
- Architecture
- Grantor
- Georgia Institute of Technology
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Alorabi, Abdulrahman Hamdan
- Advisor dc:contributor.advisor
-
- Yang, Perry Pei-Ju
- Committee members dc:contributor.committeemember
-
- Yoshida, Takahiro
- Shelden, Dennis
- Rakha, Tarek
- Balchanos, Michael
- ORCID person:identifier.orcid
Rights
- Language dc:language.iso
- English
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/81693