Generative Design

15 Jan

I first learned about genetic algorithms used in design from a lecture on design research. The name of the firm escapes me now, but they designed various interactive objects and placed them in homes to see how people live with information. One such object was a television set that picked up the radio signals of planes that passed overhead, and showed on a 3D revolving globe the flight path of that plane. The aerial for the set was designed by NASA, using a genetic algorithm to find the best possible configuration for the aerial.

Then someone showed me Theo Jansen’s strandbeests which are also designed using a genetic algorithm. Jansen had an idea of how its remarkable legs would move, but he had no idea about the best, exact lengths of the individual members. Each leg is made of 11 members, and could possible be cut in many lengths, which results in billions of possible leg configurations. Rather than test every single configuration, Jansen used a genetic algorithm to reduce the number of configurations.

“It is immediately obvious that testing every possible leg would take, well, way too long. So rather than testing all of them, a genetic algorithm tests *a few of them* over successive generations. Essentially, the computer randomly generates a hundred or so legs, tests each one of those hundred, and then rates their walking ability against each other. The computer takes the top ten performers, mutates each one of these into nine slightly different ones, and tests this following batch of legs. It repeats this process through a large number of generations, slowly refining the best performers. In this way one can find some of the best performers, without actually having to explore the entire space of possibility.” via culturalvisas

In this case I would tend to think of Jansen’s method as being halfway between a heuristic and an algorithm, since it isn’t testing each and every permutation on the list, but it isn’t using specific criteria to narrow its search either.  Rather it is checking the results against each other in order to shorten the list more quickly. This is dependent on the

At this time my friends and I were discussing the need to design new things. Questions like, “does the world need another chair?” were thick in the air at the studio, while the statement “everything has been done” could be heard from time to time. I thought that an interesting remedy to that problem, and possibly an end to product design, would be an algorithm that would generate products. Given that there are fixed values for a product, or at least a range of acceptable values, such as the dimensions and height of the seat of a chair, the angle of the back etc, then all the other values could be randomly generated to give you a wide range of possible chairs. The good results could then be cross bred, if you will, and over successive generations you could get wunderchairs.

Follow the link to a discussion on the definition of generative design, and how it’s different from parametric design. Sivam Krish discusses the theory and methods of generative design at .

I will be reading up on genetic algorithms this next month and I’ll post the tutorials and results.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: