Surrogate Models for Urban Regeneration

Goals

Conventional environmental simulation approaches in urban design are time-consuming and often incompatible with the fast-paced decision-making processes. This project aims to revolutionize this outdated paradigm by developing surrogate models that accelerate simulations (typically via machine learning), thereby offering real-time feedback to urban decision-makers, such as architects, urban designers, and policymakers. Our goal is to seamlessly integrate sustainability considerations into every step of urban decision-making processes by integrating our models with industry-standard CAD tools such as Rhino and Revit and others.

Issues Involved or Addressed

Many performance-related decisions in architectural and urban design happen too late in the decision-making process to ensure they suit the clients’ purposes. We believe this hinders true urban regeneration. This research course challenges this status quo by developing software tools that empower communities and urban decision-makers.

MacLeamy Curve illustrating the need for rapid performance feedback in architectural decision-making. (more info)
Our models will enable urban decision-making by enabling real-time testing of interventions. By involving a multitude of urban;- Community leaders- Civil society:”ƒ- local non-governmental organisations”ƒ- community-based organisations”ƒ- non-state armed actor- International actors and donors- National government, sub-national and local government- Urban planning institutions- Architects / Urban Designers- Private sector- Academia”>stakeholders, we make regenerative cities tangible, actionable, and inclusive. Our work will address:

Air quality, pollution, natural ventilation potential
Microclimate assessment (outdoor thermal comfort, urban heat islands)
Flood, stormwater, water runoff
Mobility, walkability, transportation
Global warming, climate change (heat waves)
Urban building energy, district energy, decarbonization modeling
Etc.

Partners/Sponsors

A previous software development effort by the PI (eddy3d.com) attracted 100k+ unique downloads since 2018 from within the sustainability community and is currently used by thousands of active users at leading institutions in academia and industry. That said, the models developed in this course have the opportunity to be packaged as software and to be disseminated to an existing global community of sustainability professionals.For this VIP course, we will partner with Perkins & Will Research regarding the selection of modeling topics. Every summer term, a small subset of students will have the possibility of internship opportunities at the P&W Atlanta office to take the models developed in the classroom into a professional setting and to test them on real-world projects.

Methods and Technologies

  • Surrogate models
  • Software Development
  • Validation
  • Multi-objective analysis
  • UI/UX
  • Machine Learning

Majors Sought

Computing: Computer Science, Human-Computer Interaction

Design: Architectural Technology, Architecture, Urban Design

Engineering: Civil Engineering, Industrial Engineering, Mechanical Engineering

Liberal Arts: Public Policy

Sciences: Physics

Preferred Interests and Preparation

We seek an interdisciplinary team of highly motivated students. Experience with Python, C#, JavaScript, machine learning, and simulation modeling will be advantageous. Our sub-teams typically are interdisciplinary and consist of students from:
 

Computer Science / Engineering (Civil, Environmental, Mechanical, etc.)
(Applied) Industrial and Systems Engineering
Architecture & Urban Design
City & Regional Planning
Applied Math & Physics
Public Policy
Etc.

Advisor

Patrick Kastner
Patrick Kastner
patrick.kastner@gatech.edu

Day, Time & Location

Full Team Meeting:
11:00-11:50 Thursday
Van Leer 465

Subteam meetings scheduled after classes begin.