Summer 2018 – Part 2

Our 2018 summer research internship has come to the finish line! In the last 8 weeks, the Proving Ground research team developed design uses cases for multi-objective evolutionary solvers and  created game engine prototypes to visualize complex spatial databases. The team learned a lot this summer and we wish Maren and Ian all the best as they prepare for their next academic year at the University of Nebraska and Iowa State University!

Minecart Data and Game Engines

We have long been fascinated with new ways to explore and visualize data about buildings. Our open source tools, like Conduit, aim to make data visualization easy and interactive within the design environment. Meanwhile, tools like Minecart allow us to aggregate large volumes of building data together for the purposes of data analysis exercises such as benchmarking. However, building data is more than tables and graphs: it is often essential to be able to query and interact with data in 3D to assess spatial characteristics and relationships

This summer, Ian used the Unity game engine to create interfaces for navigating our Minecart database to query and visualize spatial facility data. This data also includes mesh geometry representing space boundaries harvested from IFC files. Ian developed a 3D navigation tool that allows users to dynamically query spaces from a relational database based on different attributes and properties.

Evolutionary Computation

Evolutionary computation has origins dating back to the 1950s as a concept inspired by biological processes – such as natural selection – to ‘evolve’ solutions for complex problems. Today, evolutionary algorithms are often at the heart of digital design narratives that tout the creation of “thousands of options” or employ popular buzzwords such as “optioneering”. If you were to go by many marketing hype, it would seem as though it will soon be commonplace for a designer to simply feed constraints into a computer and then be presented a variety of valid options by way of digital alchemy.

Octopus-SolutionSpaces
Within Octopus, various design solutions are mapped onto a single graph where different axis are representative of measured outputs such as “cost” or “visibility”. Different problems will yield vastly different solution plots and patterns.

The reality of this technology is perhaps far less fanciful than the hype implies, and far more practical than it is given credit. Evolutionary algorithms are a very mature computational approach made increasingly popular by more sophisticated user-friendly interfaces. So what type of design problem makes for a good candidate to apply evolutionary computation? Three criteria immediately come to mind:

  1. A problem that can be programmatically modeled with well defined constraints
  2. A problem where solutions can be sufficiently numerically evaluated
  3. A problem that would otherwise be inefficient to solve using conventional computational or non-computational methods.

Throughout the summer, Maren explored a wide range of potential applications involving costs, sight lines, and geometric ratios. These explorations revealed some interesting design solution spaces and ultimately allowed us to better position evolutionary computation as part of Proving Ground’s problem solving methodology.

Octopus-UrbanViewGraph
This model allows building volumes to be re-positioned on site. Octopus was used to measure rays penetrating through an urban context to evaluate possible massing configurations relative to North, East, and West view cooridors.
Octopus-RiserGraph
This amphitheater model has parameters for riser height and existing aisles. Octopus was used to study how different configurations yielded better visibility performance.