Autodesk is using computational design to create more efficient products, from chairs to bikes to planes.

Using cloud computing, a program can sort through all the different ways to make a chair and find the most efficient one.

This is the type of chair computational design can generate.

An algorithm can go through a million iterations of the same chair “in the time that it would take the designer to describe just one," according to Chief Technology Officer Jeff Kowalski.

The program creates thousands of iterations of the same product based on specific design constraints, like the height and load-bearing ability of a chair.

“This applies to buildings, bridges, automotive parts--everything we do where we have to draw things,” Kowalski explains.

If the latticed aesthetic isn't what you're looking for, it can be covered by another material.

The underlying structure would still be lighter and more efficient.

It's especially useful in aerospace and aviation applications, where a few extra pounds can make a huge difference.

In the case of Moon Express, a commercial spacecraft company, computational design can come up with the best way to attach a thruster to the main body of the craft.

Co.Design

Why Algorithms Are The Next Star Designers

Rather than using software to draw and analyze a designer’s vision, Autodesk wants to use computing power to generate the idea itself. Will this render designers obsolete?

The industrial designer of the future might just be a computer. The traditional design process can be a laborious one, full of iterations and tweaks to make a product just so. For years, design software maker Autodesk has been working on an alternative: a program that sorts through all the ways to make a product of specific measurements and then spits out the best option.

"In the past couple of years, we have experienced such an explosion of computing power that we can completely change the design equation," says Autodesk chief technology officer Jeff Kowalski. Rather than using software to draw and analyze a designer’s vision, Autodesk wants to use that computing power to generate the idea itself—by running through a nearly infinite number of ways to build the same product.

Though still in the experimental phase, Autodesk has been working on computational design for the last seven years through research projects such as Project Dreamcatcher, a system that generates CAD geometry based on a list of functional requirements. This approach could prove especially effective in sectors where being light and strong are of the utmost importance—as with cars and airplanes, where extra pounds have a direct impact on fuel efficiency.

Here’s how it works. You type in a set of constraints for the product. Then an algorithm "will go through a nearly infinite set of results and head towards ones that are going to satisfy the designer," Kowalksi explains. Say you want a chair that needs to support 200 pounds of weight and be two feet off the ground while using the least amount of material possible. The program might start off with 1,000 ways to make the same chair, and then intermix what it sees as the most effective components of those 1,000 chairs to make 1,000 new options. It can go through a million iterations of the same chair "in the time that it would take the designer to describe just one." It can also suggest what type of material to use.

A computer isn't constrained by preconceived notions of what a chair should look like. The example above, chosen from several designs the algorithm generated, has a curvy, slightly portly, webbed aesthetic, quite unlike our traditional conception of how a chair should look. Other options for making the same chair might include straighter legs, or just a single pedestal leg, but the algorithm is capable of generating a variety of different designs that all meet the required criteria.

It's not a process that's limited to making chairs. "This applies to buildings, bridges, automotive parts—everything we do where we have to draw things," Kowalski explains. It could be used to design a lightweight underlying structure of, say, a bicycle, which could then be covered by a skin to give it a different look—so you're not limited to the lattice aesthetic, even if that's the most efficient material for that product.

Computational design isn’t aimed at replacing human insight. People are still better at making more subjective selections over issues such as aesthetics. But the program certainly does serve up a panoply of options and can help sort through some of the more tedious aspects of the design process.

[Images: Courtesy of Autodesk]

Add New Comment

10 Comments

  • It's been progressing to this well before computers existed. Reference standard drawings - much more rudimentary and static but a very similar concept. Anything - literally anything - a person does involves the decision making process. Mapping and then coding that process is laborious but entirely possible for entirely everything. The process of coding all human tasks will accelerate as we further progress computing that self-learns. The statement about "subjective selections" is interesting because it's currently quite easy to write predictive code that analyzes patterns, fads, movements, inconsistencies in application of preferences etc. Meaning, subjectivity can also be coded. Isn't it interesting that 20 years ago the fear was that manual labor would be marginalized in the future due to technology, specifically robotics. Yet today we're realizing that knowledge workers are actually much more vulnerable, speaking in the long term perspective.

  • Oh, well, I feel cynical today. All this indicates to me is that managers and executives or other non-designers will be running the software themselves and picking out what they like from however many iterations the system can generate for them. Why should they bother with a designer when the software can do it?

  • Hugh Tonks

    Slide 2: "Using cloud computing ..."

    That should read "Using computers ...", because using the cloud is just using someone else's computers. And I think it's pretty obvious that computers were used, so this text is unnecessary.

    Sentences need good design too.

  • "It doesn't replace people, but it could empower us significantly."

    The commenters, here, do not sound like designers; they sound like programmers.

  • As a data-obsessed web and SaaS designer, I can't help but wonder if this could be applied in the digital world. I can envision algorithms based in the following:

    • Device usage patterns
    • Navigation usage patterns
    • User role patterns
    • Language patterns (based on internal and organic search terms as well as other user interactions)

    Ideally, you'd create software which could make design recommendations based on standard usability requirements, add in your own user research, and then plug it into your web app to evaluate usage on an ongoing basis and make recommendations to humans. You could also use it to evaluate brand new interactions from humans (though of course it wouldn't replace actual user testing). You could refine it by turning some analytics options on and off. Eventually it could result in a truly antifragile UX.

    It doesn't replace people, but it could empower us significantly.

    Well, now I want to make this. Who's with me?

  • This is purely embodiment of design. Though potentially a powerful feature it is merely a tool that empowers the true industrial designer rather than replacing her.