Generative design in architecture is an iterative design process that creates unique designs basis the data and restrictions, specified by the architect. Unlike traditional design methods where the architect is aware of every possible outcome, in generative design, the creations are non-predictable. It is an exploratory process where the results can be pleasantly surprising.
Technically, to understand generative design, we must first be familiar with parametric design and computational design. Parametric design is the most basic form of algorithmic modeling. The architect specifies the constraints and features. This allows him to automate repetitive processes, by simply using a few commands. For example, using parametric design, one can define the relationship between a wall and a window, so that a change in one, reflects a change in the other.
Computational or algorithmic modeling takes this, a step further, by allowing replication of a whole set of rules and allowing interactions between them, so that entire structures are replicated. Extending our earlier example, complex relationships can be defined in computational modeling. A relationship between various rooms, entryways, and replicating this to create a building full of apartments is part of algorithmic modeling. It helps create relationships between numerous complex variables and generating designs, that are predictable.
However, in generative design, the process removes this predictability and introduces possibilities. For example, if an architect has an acre of land to work on, using his creativity and addressing constraints, he can come up with 3-4 possible designs. But generative design in architecture helps him come up with almost all possible outcomes, within constraints. It uses the computational power of computers to create various combinations using the parameters and metrics that are set. It is an iterative process and an architect has the option of filtering out & eliminating the generated designs and zero in on the ones he likes.
The process of generative design consists of the following steps:
Generate: In this step, basis specified parameters and algorithms, the AI-based system generates design options
Analyze: The system analyses the designs created as per the parameters set by the architect
Rank: The system ranks designs as per their capability of accommodating parameters
Evolve: The system shortlists preferred design paths from the top-ranked options
Explore: At this stage, the architect explores the available designs and short-lists them for future use
Integrate: The architect integrates the selected design into his over-arching project or uses them as a starting point for a new project
Generative design in architecture has shown a lot of promise since it first became mainstream. It has allowed architects to come up with thousands of design options, in a short period. This ability is especially useful in time-bound projects such as the design of a trade show venue. The question that arises here is that, do we need so many options? Well, yes.
Take the example of a trade show. When an architect sets to design it, he will use a few human rule-of-thumb or heuristics to create a design from scratch. These can be that all stalls should have visibility, should be accessible. Maybe there needs to be a central area that leads to all stalls. The constraint also matters, such as the area available and the permitted entryway depending on road access or parking area. An approach like this allows him to produce a standard trade show layout. However, generative design in architecture can produce layouts that he might have never thought of, which can deliver more value to the trade show participants as well as visitors to show.
In the world of generative design, the above is a simple problem. The scope extends to designing complex hospitals, low-income housing and other such impactful projects. The process ensures that we always have optimized outcomes.
The design process is available in existing CAD design software as a feature. Artificial intelligence and cloud computing, empower generative design. Some software that provides the see features are Siemen’s NX, Infurnia, Creo by PTC, and Autodesk’s Revit. Most users find adapting to and using generative design in architecture, an easy task.
The process generates thousands of design and such an output can easily overwhelm a user. In such cases, users can either filter out designs which do not fit or the AI-powered system can eliminate the unfeasible design. Also, the software is designed to ‘learn’ about the preferred output so that in future designs, it can recommend outputs that are closer to what the architect prefers. Artificial intelligence is used to make the software create optimized results that are efficient and practical.
Generative design in architecture shows promise in the years to come. Architecture is a creative yet logical profession, and generative design delivers both or rather more than what the human mind can fathom. We can and we should weigh the pros and cons of this iterative process in architecture. However, most architects would agree that it is a tool that certainly helps enhance one’s architectural practice and opens up a whole new world of possibilities.
