Autodesk’s Vision: Generative Design

References Cited

Mechanical Computer Aided Design (MCAD), Autodesk

Just a few short weeks ago, Autodesk held their annual AU 2014 event in Las Vegas. Mainly comprised of training classes, this event is also an opportunity for Autodesk to unveil new products and offer insight into the vision driving their products. And on Tuesday, December 2nd, they did just that.

In this post, we’ll look at the concept of Generative Design, the capabilities that Autodesk plans to provide in support of it and the implications that it has for the industry.

Generative Design and Capabilities Provided by Autodesk

So what is this thing?

Well, in the simplest terms, the concept is to enable software to design mechanical components autonomously. Yes, really.

How does it work? The process will look something like the following. The idea here is that a user determines design objectives and constraints. The software application or system then explores options to varying degrees of completeness. Resulting designs are then offered to the user to select and even refine.

OK. OK. I know. It seems pretty far fetched. But in reality, there is some precedent. Especially when you look a bit closer and see how this will conceptual work. In talking with Kevin Schnieder, the product guy for Fusion360 and Inventor, such a software tool like this will end up using many different kinds of algorithms, not just one. In one form, it might look like topology optimization, where material is removed based on loading conditions. In another form, an algorithm that mimics the growth patterns seen nature might be used. Yet another form would be some kind of design of experiments. You can mix and combine these algorithms or approaches and apply them to your design as needed.

Now, the result of such process is intriguing. By ‘growing’ such designs, you end up with very organic shapes. Shapes that the human mind could rarely achieve. But, nonetheless, these designs satisfy and often even exceed the objectives and constraints of the user.

In addition to this interesting process, capability and results, I found the implications in terms of user interface also intriguing. Here are my tweets from the event.

Thats right. Mechanical engineers might be ‘programming’ these algorithms into their designs. And, as you can see, this isn’t without precedent in the design of hardware. Electronics engineers have been ‘programming’ their PCB, FPGA and ASIC designs

Commentary and Analysis

So, what’s my take on this vision? I have a few important takeaways.

All Mechanical Engineers Need Jarvis

Have you ever watched any of the Ironman movies? In all of them Tony Stark has an Artificial Intelligence called Jarvis that does a lot of stuff. One of those things is design the suits.

Now, I’m not suggesting that mechanical engineers will end up talking their computers. But what I am suggesting is that sooner or later, all of them could have a lot of design ‘agents.’ More specifically, a single mechanical engineer might have dozens of these ‘jobs’ running simultaneously. The mechanical engineer acts like a manager, setting up the initial objectives and constraints, overseeing the ongoing work and ultimately making the final selection of a design from many different possibilities.

Why is that a big deal? I was recently talking with someone about the ratios of different kinds of engineers in the High Tech and Electronics industry. My counterpart suggested that for every one mechanical engineer, there are five electronics engineers and ten software engineers. My point here is that we’re not going to have a flood of hiring for mechanical engineers. More and more manufacturers see innovation coming from software and electronics. So with more work being dumped on the few mechanical engineers that are left, they need more and better enabling technology. The evolution of mechanical engineers into a mechanical engineering manager with many different autonomous computational agents doing the grunt design work seems like a good alternative to working seventy hour weeks.

Moving Beyond First Feasible

Lets face it though: there is more at stake here than simple productivity.

Today’s mechanical engineers are flat out busy. That means when they develop a design that satisfies the constraints, they move on. The ‘First Feasible’ approach isn’t desirable, but is necessary given current workloads. An intriguing aspect of this Generative Design concept is the possibility to greatly improve the quality of volumes of designs without dramatically increasing the workload on mechanical engineers, making it truly possible.

The advantage, of course, will be reaped by manufacturers. With better designs, failure rates drop, materials costs go down and the amount of change orders decrease.

Business Problems, Not Technology Problems

Last, but not least, I find this vision intriguing because of the type of problem it solves.

For the past five years, software providers of Mechanical CAD applications have focused on solving issues that have cropped up because of the technology itself. Better interoperability merely addresses how poorly design data can be translated between CAD applications, which is a non-value add activity in design. Improved 2D sketching tools offers gains in productivity of how to edit lines and arcs, again an issue that rose because of how poorly these tools supported concept design. Model-Based Enterprise improvements in Mechanical CAD is an attempt to evolve from the inherent inadequacies of 2D drawings.

Generative Design, whether it comes from Autodesk or another software providers, is different. It is technology applied to the improve of how products are designed, not how to use the technology. That’s a refreshing difference.


  • At AU2014, Autodesk introduced a vision to provide capabilities to support Generative Design.
  • The concept behind Generative Design is to enable software to design mechanical components autonomously.
  • Autodesk’s solution will likely provide many different ‘algorithms’ to achieve design objectives and constraints. Using such a solution will likely look a lot programming.
  • Once more widespread, Mechanical Engineers could evolve to manage many different ‘agents’, which would be different autonomous design jobs, running simultaneously.
  • This solution could improve the quality of designs, allowing engineers to move past a ‘first feasible’ approach.
  • Generative Design seems to solve a business problem, not merely a technology problem. That’s a refreshing change.

Well folks, that’s my take. Let me know your thoughts.

Take care. Talk soon. Thanks for reading.


When it comes to the value of Model-Based Definition initiatives, there’s lots of heresay and little evidence. That’s why Lifecycle Insights conducted a study and published the results in a research report.

Did you find this article valuable?
Join our newsletter community for more.

Chad Jackson is an Industry Analyst at Lifecycle Insights and publisher of the engineering-matters blog. With more than 15 years of industry experience, Chad covers career, managerial and technology topics in engineering. For more details, visit his profile.