To start with, my contention is that architectural design would benefit from a practical application of evolution. Plenty of building designs have been generated with a very conscious incorporation of automated evolution, but I have as of yet to find genuinely useful examples rather than something that's just coincidentally resulted in a pretty form. What is generally lacking in the explanations of such "evolved" buildings is what the criteria for survival were, what the varied genes were, and how many generations deep the evolution was carried out.
For a long time, I've been interested in the application of evolution in fields other than biology, used with a particular view to solve various complex problems in incredibly simple - though almost unintelligible - ways. For instance, one of the non-fiction chapters of the first "Science of the Discworld" novel (by Terry Pratchett, Ian Stewart, and Jack Cohen) discusses an exploration of evolution through a genetic algorithm approach to making an electronic circuit able to distinguish between two tones. The circuit's logic gates were the genes that were randomly varied and inherited from generation to generation, and the survivors were selected on their capacity to give a different output - 1 or 0 - to each of the two different tones, not caring which tone was given which value as long as there was a difference.
Early on, the circuit's capacity to tell the difference was non-existent or negligible. However, after sufficiently many generations, reliable differences began to arise. After only 4000 generations, the circuit would get the tones wrong barely 1 in 1000 times. At 8000 generations, there were no errors in tone distinction that were encountered. The resulting circuit, however, was very complicated and hard to understand - for example, a portion of the circuit was found to not be connected to anything else, but if removed always caused the circuit to stop working.
In my opinion, the most important part of the experiment, however, was the efficiency and elegance of solution that results from the use of appropriately constrained and defined evolution. The evolved circuit was far smaller (ie, had far fewer logic gates) than other circuits previously made to tell the difference between two tones.
After doing some Googling, I've found a few interesting sources that I'll look into over the holidays. They are:
- An ecomorphic theatre as a case study for embodied design (paper located here): mentions some interesting historical precedent to generative design of architecture.
- An Evolutionary Architecture (version of book released online located here): Covers some interesting concepts to do with the kinds of forms that can be generated, and some ways of using the internet to expose a 3D model to genetic variation.
- Autotechtonica.org (link located here): Is currently under construction, but seems to offer a few simple existing neologisms and their definitions, which might be handy to glance over if you're trying to learn about the topic like I am.
- Morphogenesis of Spatial Configurations (link located here): Talks about evolution when selecting forms based on building performance criteria such as structure and accessibility.
From the second source, I found an example of what I was talking about at the start of my post when I said "I have as of yet to find genuinely useful examples rather than something that's just coincidentally resulted in a pretty form".
An animated example of co-operative evolution by a network of computers. Pretty, but lacks useful information to explain what it is. (from http://www.aaschool.ac.uk/publications/ea/intro.html)
What I would like to create using the process of evolution for this masters studio is something of utility. I want to produce an intelligible analysis that can be clearly and specifically used to inform an architectural design. Ideally, the evolution will be applied to an area that doesn't already have its own simple solutions. I think it would be more exciting if it were applied to, for example, the problem of space organisation and linkage, which I have observed is often a point of unfounded contention between designers - eg, arguments about one space not being suited to be connected to another, and so forth.
Furthermore, it seems that evolution of useful aspects to a building would best be used as a design informant rather than a means of producing the end design in itself, since there are philosophical aspects of design that haven't yet been accurately encapsulated in formal systems (such as those used to found computer science and information technology). To clarify: I trust that sufficiently many generations of rigorously managed evolution would produce an effectively failsafe product, but if and only if the appropriate conditions of selection and the appropriate genes were known and also formally encoded in the process of evolution. However, culture and many philosophies have as of yet to be formally encoded in such a comprehensive manner. Thus, I would use evolution to inform my design process for those conditions of selection and genes which are known, but I would want to refine the design personally to ensure it suits the formally undefined constituents of philosophy and culture.
Something that seems a little more promising than the above animation is the paper on Morphogenesis of Spatial Configurations. Referring to the below image of a 3D model generated from Lindenmayer Systems (aka L-Systems) and genetic programming, it certainly seems to produce what looks like a far more sensible form, though I am unsure of what the original L-System's configuration was.
IMAGE_DESCRIPTION (image from http://technozoic.blogspot.com/2009/11/gennaro-senatore-morphogenesis-of.html)
I'm really liking how many freely available online resources are turning up for this topic. There'll be a lot of reading to do, but I suspect I'll learn a lot in the process, which will hopefully save time when it comes to producing my own evolved design informant.
No comments:
Post a Comment