Title:
Rational decision making in early urban design based on uncertain performance predictions

Thumbnail Image
Author(s)
Hashim, Alya Abdul Sattar Yaqoob
Authors
Advisor(s)
Augenbroe, Godfried
Rakha, Tarek
Advisor(s)
Editor(s)
Associated Organization(s)
Organizational Unit
Organizational Unit
Supplementary to
Abstract
The world is currently undergoing the largest wave of urban growth in human history. More than half of the global population is now concentrated in urban areas, and by 2060 `two third of the expected population of 10 billion will live in cities. While accommodate this tremendous growth, reducing urban energy consumption of resilient and livable cities should be seen as associated priorities. Meeting these priorities head on requires complex decision-making at the early phase of urban design, when a large number of parameters are still undecided, and their level of uncertainty is high. The thesis proposes a rational decision framework that responds to these challenges for a specific set of measures within the following limited scope: energy efficiency in urban layout, indoor daylight level, network connectivity, outdoor public space visibility and thermal comfort. The early stage of urban design is characterized by its iterative nature of repeated alternative generation (divergent phase), and alternative assessment and selection (convergent phase). Decision making occurs during or at the end of these phases with considerable uncertainty in the many as yet unresolved design parameters. Therefore, methods and tools applied during these phases should account for the iterative and unpredictable nature of later design evolution. Currently there is no consistent support for rational decision making at the early stage of urban design. Typically, single deterministic predictions are generated based on assumed parameter values when in fact many of those parameters have not been decided yet. This dissertation starts from a hierarchical structure that outlines consecutive steps in the design process by geometric output type. This is not the main focus of the thesis but merely a structuring principle that is employed by the rational decision framework. This framework supports the comparative assessment of competing design alternatives under uncertainty. This is the main focus of the research. It introduces explicit information about uncertainty in undecided design parameters and analyzes their effects on the confidence with which one design variant can be prioritized over another. The approach is implemented in a Rhino-Grasshopper platform for five concrete performance measures: network connectivity, visibility in open space, outdoor thermal comfort, building energy consumption and daylight utilization. Low-resolution simulation models are developed for each of these measures to service the iterative nature of design with fast computation of results. The resulting models serve as normative substitutes for more accurate physics-based prediction models. The research has developed a systematic verification approach showing when these reduced order models are indeed as adequate for the targeted comparative analysis in early design as their high-fidelity counterparts. In the comparative analyses of design variants, point values of inputs are replaced with probability distributions that quantify the expected variability (treated as design uncertainty) in later decided design variables using a Monte Carlo technique. Hence each generated outcome is a probability distribution that represents the uncertainty in the performance prediction of a design alternative under study. The performance predictions are the inputs into the decision making allowing the designer to make a rational choice of one design alternative over a competing one. In the developed framework such decisions are driven by minimum required confidence levels that a decision maker is comfortable with when prioritizing a variant. As an associated issue the research tested the effectiveness of current rules of thumb and found that design choices that they suggest typically fall short of the confidence level required by the decision maker. This dissertation introduces the methodology, the development of a framework for comparative analysis with embedded normative models (implemented as grasshopper components) and their execution in the current prototype.
Sponsor
Date Issued
2020-03-06
Extent
Resource Type
Text
Resource Subtype
Dissertation
Rights Statement
Rights URI