Generative design and performance modeling for relationships between urban built forms, sky opening, solar radiation and energy

Abstract

For the past two decades, extensive research has been conducted to evaluate the performance of urban form scenarios such as sky view factors, solar radiation, or energy performance. Simulation-based performance analysis has been applied for optimizing urban design using multiple performance criteria of energy demands, daylighting, or radiance. Two major limitations exist in the performance evaluation methods of urban design: 1) the number of urban design alternatives or scenarios has been limited, and 2) various performance criteria and metrics have not been assessed synthetically. In this respect, we aim to generate all possible alternatives under constraints such as a city policy or development requirements. This paper addresses the integration of generative design and multi-criteria performance analysis for urban design decision making. The relationships between geometric urban forms and multiple performance criteria are identified. The proposed methodology incorporates the reinforcement learning algorithm, an advanced machine learning technique, along with parametric modeling techniques, to enhance urban design decisions based on various performance criteria. This methodology was tested on the design of an international campus of the Georgia Institute of Technology, located in Shenzhen, China. An algorithm generates numerous geometrical alternatives of campus design under site constraints and development requirements of the campus. Those alternatives are then evaluated based on performance criteria of sky openings, solar radiation, or energy consumption synthetically. The method enables an integrated performance analysis for all possible design under constraints. The paper proposes a data driven urban design approach that connect generative design and multi-criteria performance analyses. The relationships between urban geometric forms and performance criteria function derive guidelines for a sustainable and green campus. In future conduct multivariate analysis to examine the relationships between geometric elements and performance measurement metrics.