Data Driven Architecture
One of the largest resources we have today is data. We have information on almost every measurable subject. But what do we do with it? This project explores how data can directly influence architectural form. The results of this exploration into data driven architecture is that of an automated system that based on its location, users, and surrounding data can produce an efficient and performative building. Not necessarily performative in purely the ecological sense but rather performative in all measured ways to satisfaction, circulation efficiency, types of program,
This project focuses on a site in downtown Chicago. It lies in a dense area populated by residential towers and a few office buildings. The program for this building is for a base that serves as a hub for the tower’s residents and the surrounding 20,000 residents.
The tower would serve as residences. This resulted in the need of two frameworks one for the retail base and one for the residential tower.
This project aimed to find a direct connection between data and building and how the two can interact. This resulted in a revision to the definition of an architect as one who designs frameworks rather than buildings. A Framework is defined as a basic structure underlying a system. This definition can be applied to architecture in that rather than producing a form as a result, architects produce a framework or basic structure and allow context and data to drive the actual form working within the given framework. An analogy that works well with this concept is that of a automobile. All automobiles start with a basic unchanging framework that includes steering, wheels, and axles. It also has a specific size limitation (depending on the regulation for that type of vehicle). All of these factors could be considered a framework, or something to work within. It allows for changes while maintaining basic elements. Vehicles forms are then produced based on user needs, market demand and other data. So, while the basic framework remains unchanged, the form and shape of the vehicle can vary quite drastically.
Residential towers seem to grow anywhere and everywhere in an urban context. Stacked floors with similar floor plans for every unit. But what if you could layout your own unit? What if you paid for exactly what you wanted and nothing else? The goal was to develop a system which assigned cost to square footage, views, elevator proximity, height, and other parameters, and allowed its users to lay out their unit based on a predefined 12×12 module system. The system would then take all of the units of the tower and optimize their location so everyone’s priorities are met. The first step was thus establishing the limits. Each floor has 8 units: 4 corner units and 4 standard units. Each unit begins with a utilities module which contains all of the utilities for the unit and the entrance to the unit. Vertically, this module is in the same location for all units allowing for easy vertical transport of HVAC and other utilities. 5 other modules make up the base of each unit as this is the minimum size that is allowed. This is equivalent to 864 sqft. The actual process of laying out the unit is where this system comes into play. An app was developed so users could do this. The app begins with collecting information about the user and what their priorities are. This includes showing the user views from different sides and heights of the building and asking their opinion, finding how important a quick elevator ride is to you, and many others. Since you are not buying a specific unit on the tower, this first portion of the app is meant to find the users overall preferences when it comes to their unit location on the tower.
The second portion of the app is a visual layout of a typical unit (either corner or standard depending on which you prefer). Here the user begins to program their unit by placing living rooms, dining rooms, kitchens, bedrooms, patio space, etc. and choosing their sizes. As stated before users begin with 6 modules (5 open ones and 1 unchanging utility module). They can then move outward depending on how much space they need. While all of these design decisions are being made in the app, the program is keeping a running total of cost which is assigned to all aspects of this process. For example, if a user wants to move outward 4 modules they can, but the further they move out, the more expensive each square foot becomes. This is to cover the additional cost for structure that is a result of the cantilever. The running total is always in view of the user so they can keep tabs on their cost and make sure they are staying within their budget.
The question then becomes how does one interact with their neighbors? If you are choosing to move outward from the tower but your neighbor wants a view in your direction, how is this negotiated? When the user specifies a priority that may come in conflict with a neighbors, a max bid is request from the user. When placing the unit on the tower, the programs number one goal is satisfying every users priority. But while this is sufficient in 90% of the cases, it isn’t possible for all of them. Therefore the next override is the bid. The User willing to pay the most takes priority.
Once all of the units for the tower are laid out and sold, the tower begins its optimization process. This process works by placing all units at a default location. Each priority given by the user is tested and given a score. The units are then moved randomly at first to give the program an idea of the possible solutions. Through an evolutionary optimization process the units are moved, priorities tested, and scored then repeated. This continues until the program reaches a solution it believes scores the best possible. This becomes the towers final geometry and the construction process can begin.
Fusing these two frameworks delivers a unique building tightly and directly connected to its context and to its users. This program can take any given location and data from users and lay out a unique form based directly on its context and its users in a matter of minutes. Therefore this same system could be applied with minimal effort anywhere in the world and produce a unique contextual product.
The program for any retail environment is tied mostly to its context. Who will use your space? What is near your space? How do people get to your space. All of these questions go into the decision for program size, type, etc. This is typically done manually. The goal here was to develop a system where this is done automatically. Based on a single locational point and parameters for size limitations, program preferences, etc. The process begins calculating.
It starts by first gathering input about its potential users by way of mobile apps that survey for answers to questions. The basics of these questions ask surrounding users how far they are willing to walk for specific programs. For example, one question asks users how far they are willing to walk for a bite to eat. The user answers in number of minutes. The answers to these questions are instantly synced with a online database and thus begins the calculation process. Once answers start flowing in the program begins by getting walking directions from the given site to all surrounding buildings. The distance is calculated and compared to the answers to the surveys. If the survey says people are willing to walk 4 minutes with a load of groceries, then only buildings within a 4 minute walk are included in the results. After determining which buildings are within walking distance for the given program, the population of those buildings are determined using GIS information including square footage, number of units, etc. Adding these determines the total customer base for the given program. Additionally according to each building’s location the census data for each is queried and downloaded. Returning detailed demographic data to be used later. The result is an accurate picture of each program’s customers and their demographics. Each program customer base is different because people are willing to walk longer for certain programs than for others. For example a person may only be willing to walk 4 minutes to grab a load of groceries but would walk 9 minutes to get a bite to eat. Thus the number of buildings and therefore people, is less for the grocery program than for the restaurant program.
From each programs industry leaders a sqft/capita is determined (ex. National Grocers association). Using this, combined with the customer population a desired square footage can be determined.
Next the program uses Google to search the area for similar programs and uses a GIS database to find their sizes. Their sizes are then subtracted from the desired square footage for the area and the result is the additional square feet that location would be able to support for the specific program.
Using a combination of this square footage data and the demographics of the programs customers, the application begins to find a good fit for the space. For example, if an area could support 18000 sqft of grocery space, and the customers have an above average income, a Trader Joe’s would be a good fit. This is because the average size of a Trader Joe’s store is around 16000 sqft. The stores also provide a higher quality product but at a higher price. Alternatively if the customers made a relatively low income, a urban Walmart may be a better fit. The application pulls these stores from a database ranking each on quality, cost, and size. Demographics like income, family size, gender, and others are considered by the application and the decision made. Then considering site constraints, the programs are laid out. The application considers points of access to the site and other factors when optimizing the locations of different programs. One of the main organizing factors that the application considers is a spaces relative size and its proximity to an anchor space. Anchor spaces are places that draw people in like a grocery store, a nice restaurant or a well known retail space. The application places smaller spaces at prime locations between where users enter the building, and the anchor spaces.