Autodesk’s generative design tool came up with over 10,000 layout possibilities based on design goals. From there, it measured how each design performed and picked the best. (Image courtesy of Autodesk.)
You get two, maybe three layout options and each appears as if it ignored your inputs more than the last.
Your office layout can make or break your company’s productivity, morale and ergonomics. Yet we never seem to spend that much time optimizing these spaces.
Autodesk knew that with the opening of its new office in Toronto, Canada, it could do better. The result was its generative design tool, which can be big news, not just for BIM, but for CAD, CAE and many other design tools.
Don’t believe me? Look at the results from Autodesk’s internal pilot project.
“Our project was to create computer code, take information from all concerned parties, synthesize that feedback into requirements for algorithms, and generate office layouts,” said Gordon Kurtenbach, head of Research at Autodesk.
At the very first meeting with Autodesk’s research team, the team gave Kurtenbach eight basic office designs for the space. He said, “That’s great—this is much more than the three I’m used to getting!”
“No, no, no,” replied the research team, “These are just representations of the family of designs. We generated 10,000 possibilities and selected these from that.”
An architect’s job doesn’t change much between the two meeting examples. They still need to recognize what the customer needed, code that into industrial design concepts, and then come up with the best designs to meet the customer’s criteria. The software merely augments this process by coming up with thousands of design options for the architect to choose from, rather than the architect coming up with a handful of options.
Now, multiply this concept to the design process in general. Think of what could be generated.
How to Bring Generative Design to Layouts and Assemblies
Genetic algorithms don’t care what they are optimizing. They just optimize based on the parameters and preferences of the users. This means that Autodesk’s new tool has the potential to help out countless designs and layouts in various industries and job roles. (Image courtesy of Autodesk.)
David Benjamin is the head of The Living at Autodesk, and his team served as the interior design architects of Autodesk’s new office.
Benjamin’s team also brought the generative design tool to life using genetic algorithms, which gave the program the flexibility to work with many parameters, goals and applications.
Genetic algorithms define the steps Autodesks' generative design tool uses to come up with the next design. The algorithm is based off of natural selection, hence the name.
Benjamin notes that this new generative tool isn’t just for industrial designers. Manufacturing engineers, plant designers and even mechanical engineers can theoretically use this technology to optimize their workspaces or designs.
“It’s flexible and able to account for 6, or even 10 goals, for the space,” confirmed Benjamin.“It’s a good platform to work for a specific space, or a part in aerospace. Say you want to optimize a part’s weight and structural performance. This was a goal we had for Airbus to make a lighter-weight aircraft part. We’ve also had goals to optimize daylight and productivity in a workspace like Autodesk’s Toronto office. We can use the same genetic generative design algorithms.”
Think about it. How much difference is there between optimizing the layout of an office and optimizing a factory, or even optimizing a mechanism? It’s all a bunch of parts that need to be fit and arranged together in the best way possible. What changes in these scenarios are the parameters and constraints that govern the placement of each part of your puzzle. The computer doesn’t care what those parameter or pieces are; it just uses the algorithm to optimize them to the programmer’s tastes.
“We like using genetic algorithms as it’s a very flexible approach to creating a lot of designs and learning over times which are better and which are worse,” said Benjamin.“It’s also a good algorithm for exploring very widely and not to quickly home in on a design you think is good. It still explores all kinds of possibilities.”
How Autodesk Used Generative Design to Optimize Its Office Space
Comparison of various spaces and neighborhoods in Autodesk’s new Toronto office. (Images courtesy of Autodesk.)
Autodesk chose to test constraints associated with optimizing their interior work environment in the Toronto office.
The constraints were things like load bearing walls, elevator shafts, hallways run by the building management, and HVAC systems. The parameters governing the placements of each office part were:
- Adjacency preference—a score of how near a desk is to things the user wants to be near
- Work style preference—a measurement comparing a team’s space with their wants (e.g., Is it quiet or more social and which do they prefer?)
- Buzz—a measurement of how spread out high activity zones are to maximize their use
- Productivity—a measurement of how distracting an area might be
- Daylight—a measurement of how much natural light an area receives
- Views of outside—a measurement of how much a desk can see out of nearby windows
“These are the goals we looked into, but we also explored other goals,” said Benjamin. “Each generative design system had a slightly different set of design goals and geometry constraints.… Each system came with 10,000 design options.… That was a fascinating part of the design process.… How do you translate what you want out of the project into a set of measurable goals? What is included in generative design?”
The Autodesk office houses many different teams, each of whom would have their own preferences for their work area. These work areas are dubbed "neighborhoods." The boundaries of these neighborhoods could be moved to a certain extent, but were really used to guide the software to make working spaces for each team.
As each neighborhood would house a team, each team had to codify its preference for its office in various surveys.
Each neighborhood in the office has a different look and feel. A team with a need to engage in a lot of collaboration would likely end up with a neighborhood that has a more open concept. A team that is heads down on details will likely see more cubicles or office doors in its neighborhood. A team that works on testing and building physical equipment will likely see more of a workshop space than an office. And, of course, that workshop neighborhood will likely be located closer to the collaborative team’s neighborhood than the heads-down team, which needs a quiet space, after all.
“We had some teams or employees ask, ‘Why is my team here?’ when they don’t get a lot of sunlight,” said Ramtin Attar, head of Design and Social Impact at Autodesk. “It’s because your team said on the survey they preferred a dark or quiet space. They didn’t realize how much it affected the results. Some even asked to change their inputs.”
When Will Generative Design Come to Autodesk Users?
Autodesk’s generative design tool has a lot of potential for much of its portfolio of tools. (Image courtesy of Autodesk.)
One of the reasons why rule-of-thumb and the-way-we-have-always-done-it design remains so strong is because of the mentality that if it worked the first time, why risk it? This becomes more of an issue as the complexity of a problem grows.
“There is no reason why the complexity of thousands or hundreds of thousands of parts can’t be encapsulated in something like generative design,” said Benjamin.
Factory layouts, office spaces, jet engine assemblies, city neighborhoods—this tool can be used to optimize them all. The question is, when will we see it? Will it be a stand-alone or plug-in tool for any of the numerous CAD, CAE, BIM, CAM or other design software offered by Autodesk?
The possibilities for Autodesk to flip this concept into one or even multiple tools is exciting to Andrew Anagnost, CEO of Autodesk. He sees it as an opportunity to save many organizations from the wave of disruptive technologies affecting various industries.
“Almost every customer we deal with is being disrupted in some way,” said Anagnost.“Our customers are seeing massive changes in the way they do things. For example, in the AEC space, construction companies are trying to figure out how to act more like manufacturing companies. The companies that win in the future will do more with less—less material, less resources, less impact in the surrounding environment and impact with local municipality with regulation and building codes.”
Given Benjamin's statements on the flexibility of generative design tools that are governed by genetic algorithms, it sounds like it can handle optimizing everything from building codes, regulations, materials, resources and environmental impacts.
“Autodesk and AI understand geometry,” agreed Kurtenbach. “We’ve done CAD for years. Banks make AI to understand financial data; we can use it to understand shapes. We have a lot of models and objects to help the computer understand what a bolt is. But there is also non-geometry stuff that relates to processes that our customers want. There is an automation part, and machine learning can be used to solve a lot of problems.”
So, we know it can be used in much of the Autodesk portfolio, but when will the company actually roll out something? For that, Autodesk says to keep your eyes open.