Part of a building designer’s daily work is to draw and redraw lines, shapes, objects, and images. From the ‘starting line’ to the ‘finish line,’ there’s a redo for every do. From efficient use of space to furniture placement, every design choice is crucial to the ultimate success of the finished product.
It’s never an easy process—but through the devices of Generative Design, said process has been rendered a bit easier and more enjoyable.
Generative Design intermingles parametric design and artificial intelligence with the restrictions and data introduced by the designer. Celestino Soddu, a researcher at the Politecnico di Milano, defines generative design as a morphogenetic process that incorporates structured algorithms like non-linear systems to achieve one of a kind and unrepeatable results, executed via an idea code, as is discovered in nature.
For example; when one examines the design of a tree, a large trunk wider and stronger at the base can withstand the pressure and tension brought about by the wind and its own weight. From that point, a number of thinner branches emerge, with everything coming together in the leaves. No materials are left over, and the forms assumed are the best suited for their habitat. In windy locales, the composition of the tree will differ from that found in a sandy terrain, distinguished by a process of natural selection taking place throughout millions of years. This same principle can be applied to art, design, and architecture.
The idea is based on a thorough investigation of design options, taken from specific assumptions made by the designer for a certain purpose. Generative design is a design plan that increases human resources through the use of algorithms designed to automate design logic. The designer still sets the parameters, but as opposed to modeling individual elements, generative design software helps you devise a multitude of solutions at one time and at times even find unexpected solutions not attained through conventional means.
Brazilian architect Guto Requena employed generative design to devise stools whose volumes were formed by the rhythm and melody of popular Brazilian music. This inspired the creation of organic shapes cut into pieces of marble. In the Netherlands, startup MX3D united with Laarman Lab, Heijmans, Autodesk and other supporters to make a pedestrian bridge produced with 3D printed steel. The team used generative algorithms as tools to come up with successive design iterations in accordance with a certain set of parameters. After deciding on a shape, digital simulations of the bridge were set into motion, stripping away excess material by intermingling structural calculations with geometric manipulation, instructing the algorithm to know which parts of the bridge were least important. The project employed Generative Design to mix the potentials of the machine’s 3D printer with different tests of shapes and design possibilities through the application of minimal parameters.
Also relevant are the results of the research project Evolving Floor Plans, which investigates speculative and optimised architectural organisations through the use of generative design. The images of classrooms and the movement of students in a hypothetical school were produced through a genetic algorithm programmed to cut down on walking time, hallway traffic, and other considerations. The floor plan grows out of genetic coding through the use of indirect methodologies, like contracting graphs and growing corridors, modeled on an algorithm inspired by the design of ant colonies.
Of course, generative design will not always produce complicated and organic shapes. It can produce the repetitive and uniform design processes that we recognise—but with a beautiful, creative twist. In 2019, Sidewalk Labs presented a generative design tool that applies machine learning and computer design to the cultivation of urban planning scenarios. Applying geographic data, urban guidelines and rules, street grids, orientation, weather patterns and structural heights as input info, the tool creates potential scenarios for building designers and planners to consider and modify their final product. Via machine learning, the system has the capability to enhance the design task and produce improved projects as it is used more and more.
Generative Design played a vital role in the design of the Autodesk offices in Toronto. The process was launched through the collection of opinions from employees and managers regarding work styles and location specifications, which were morphed into data. As a result, six primary measurable parameters were determined:
Adjacency preference: Reduce the distance standing between workers and office amenities;
Work style: Determine the ideal location for each team and determine their light preferences and noise levels;
Interconnectivity: Activate and make the most of commonly shared spaces;
Productivity: Lessen visual and audio distractions;
Certain elements could not be changed, like vertical circulation, restrooms and plumbing, and the building structure. Negotiating these obstacles, the design procedure was automated to consider thousands of layout configurations from hundreds of blended variables, attaining the performance classification of each alternative for the specified parameters. Once the space is reoccupied, and the productivity or frequency of the use of each space is recorded, it is feasible to cement or alter certain parameters to render the model more workable for this or other projects. If any parameters shift, like an increase or decrease of the amount of teams or if a new space is required, these factors can be included in the database to design new design versions.
In a project centred around the idea of Generative Design, the computer is no longer just a graphing tool, or an online storehouse to register materials and dimensions. This form of design empowers the computer to become a collaborator in any design project.