What Can We Learn from Games?: Goodbye Parametric Design, Hello Procedural Generation

Synopsis:

The AEC industry has to a large extent embraced parametric design. Will procedural generation, as distinct from parametric design, gain similar traction in the coming years? If so, what are the benefits of ‘procgen’ and what can be learnt from the games industry? 

Learning Objectives:

1. Attendants will develop an understanding of what procedural generation is, how it differs to parametric design, they will also gain an appreciation for the potential benefits to the AEC industry, and learn some key terminology to be able to discuss procedural generation with colleagues and collaborators. 

2. Attendants of the talk will become familiar with a number of procedural generation algorithms, how they work and what kind of problems they can be applied to. Attendants will also gain knowledge of how to explore this topic further, e.g. useful resources and keywords to use in searches. 

3. Attendants will gain an appreciation for how to approach generative design problems. This will be explained through the use of a case study, sharing lessons learnt, how to simplify a problem, and where the game analogy falls short for real life AEC problems. 

Body:

Computational design is filtering down from use on ‘starchitecture’ projects to more conventional projects. Many parametric efforts are in the form of bespoke scripts with hard coded logic which limits the scope to a particular project. This talk, however focuses on procedural generation, where solutions emerge with potentially unexpected results. One related example in the AEC industry is topology optimisation used to create structural forms. This talk describes general procedural generation techniques which can be adapted to many varied applications. In particular we will discuss constrained geometric layout planning whilst maintaining a diversity of solutions.

Games have been using ‘procgen’ at least since the 1980’s to generate game levels and other content. This presentation asks what can be learnt from this rich history in applying procgen to AEC applications. The benefits of looking to an industry with a history of procgen include: developing a vocabulary to discuss the topic; gaining an appreciation for various algorithms, old and new; and an understanding for the limitations of procgen.

Concepts and algorithms that will be covered:

The “1,000 bowls of oatmeal” problem

Monte Carlo Tree Search (MCTS)

Space colonization algorithm

Constraint satisfaction problems

As a case study, the talk will delve into how a procedural generation was used in developing the Modulous platform. A web app which digitises design, costing, programming, and procurement for the design and delivery of affordable and sustainable homes. The goal is to help eliminate the housing shortage in a sustainable way aiming for net zero by 2030. The thought processes and lessons learnt throughout developing the application will be discussed, to help inspire others in their procgen endeavours. ”