From CAD to Machines as Co-Designers

modern parametric

Finnish economics professor Osmo A. Wiio claimed that we typically overestimate the near future and underestimate the distant future. His thoughts resonate well with today’s perception of design automation.

Back in the 1980s, I took part in the first-ever Finnish “integrated CAD” projects, in which the architect, the structural, HVAC, and electrical engineers exchanged CAD files instead of blueprints. We used minicomputers, which had about as much computing power as today’s laser printer. Still, we were able to automate design tasks.

Our flagship project was the biggest office building in the Nordics. The design consisted of repetitive floor units and, thus, was ideal for primitive design automation. We could, for example, make changes to a few base units and propagate the updates throughout the floors and blocks of the building. It was a huge time saver if you compare it to manually doing the same operation a number of times to dozens of drawings.

Today’s designers have tools that can do much more and much faster. New technologies don’t just speed up design tasks, they can also create completely new value to clients, designers, contractors, and manufacturers.

Parametric Design

Our 80’s automation could be called scripting – routines that use a series of commands to execute a task. Today, we have a plethora of concepts that relate to design automation. They include parametric, associative, algorithmic, computational, generative design, and others.

The common thread among the new concepts is the use of parameters to define the features of a design object. Parameters are typically dimensions, quantities, and other physical attributes. Parameters could also be related to costs, time, and material qualities. Parameters can have constraints and associations with each other.

The power of parametric design becomes obvious with the use of digital tools that allow us to explore geometries using algorithms. Repeatable, fairly simple rules that use parameter values for input to modify the geometry of a structure can create unforeseen forms. Algorithmic design has other uses in designing floor layouts, facades, and even city plans.

Parametric design, Zaha Hadid Architects

Shajay Bhooshan, Senior Associate at Zaha Hadid Architects, told me in an interview that computational design is not just about architectural forms: “We believe that modern, intricate life requires more complex design. Our environments need to adapt to fast-paced change, and parametric design makes it possible. We want the built environment to mirror the natural environment.”

Parametric design goes hand in hand with digital fabrication. Computational design and robotics make mass customization feasible. A robot will make one or one hundred copies of the same object without additional costs.

Generative Design

What happens if you let the computer change the values of parameters of the algorithmic model by itself? It can test out thousands of parameter combinations in a very short time.

The generative design approach creates a set of alternatives that you can evaluate in terms of tonnage, piece counts, areas, costs, heat conservation, and so on.

Joel Simon’s experimental research project, called Evolving Floor Plans, explored speculative, optimized floor plan layouts. He used the space layout of an elementary school as a starting point. He ran it through his generative design software that aimed to, for example, minimize the traffic flow between classes or minimize the fire escape paths.

The result was a layout that looks “organic” – totally different than the original rectangular design. According to, Simon claims that his tool could be used to “breed” buildings, with the use of digital technologies, such as 3D printing.

Adding Machine Learning

Artificial intelligence and machine learning have been in the headlines, but, so far, we’ve not seen commercial breakthroughs in the construction industry. Coupled with parametric and generative design, they will, ultimately, create a totally new playground for built environment designers, builders, and owners.

Machines can use data from existing designs, live data from buildings and environments, and biometrical and behavioral data from people in the built environment.

Artificial intelligence will help with optimization, but also in automating rote work. In 2018, I interviewed Sweco’s Mauri Laasonen and Ricardo Farinha. They experimented with machine learning in the design of a small power plant. After the learning process, the machine was capable of automatically designing the joints of the plant’s structural model. It designed 77% of the connections successfully without human intervention.

Can bots replace a human designer? I’m confident that it will happen in some specific areas of design and construction.

More Choice

In the 1980s, one of my older colleagues taught that to please the client, the architect should present three alternative initial designs. One of them should be flawed, one should be OK, and one should be magnificent. The architect should then play their cards so that the client chooses the last one. Unfortunately, that did not always happen.

Computational design can offer infinite alternatives at the same cost as traditional, standardized production. The design team can use a large number of factors to evaluate the lifecycle performance of the alternatives. Feedback data can further improve the design.

In the 1980s, many architects feared that the use of CAD would lead to unimaginative, mechanical architecture. Looking back, our tools back then did not lead to a certain kind of architecture. It was the other way around. The architecture allowed us to use technology efficiently. Today’s computational design, however, radically changes architectural expression.

The Future

If used wisely, design automation and AI should lead to better buildings and environments. Designers can better take into account the increasing number of factors that impact the investment cost and the life cycle value of their design.

Machine learning from human behavioral and sensor data will provide designers with feedback that improves future designs. Clients can make informed decisions based on a rich set of data.

I think that computational design will ultimately change the professions. Design disciplines will amalgamate, and design and manufacturing will be tightly connected. Designers will have to consider and understand manufacturing better than they do today.

A FEM model
Ains Group: A FEM model

Computational design will spread out to areas beyond special structures and unique buildings. High-volume housing is especially ripe for design and construction automation.

My prediction is that we’re going to see “meta-designers”. They are teams of architects, engineers, programmers, economists, behavioral specialists and other professionals who develop the design process and automation of a specific project or building type. AI-driven algorithmic systems take care of much of the traditional designer’s work.

I concur with Shajay Bhooshan’s thoughts: “Architecture does not exist in a bubble. Robotics, 3D printing, AI, big data, and so on will have an impact on the industry. Design cannot be simply a matter of intuition when you’re solving complex problems.”

Share This