The system is a combination of search and optimization algorithms and computational geometry which is developed as a data-driven approach with input from people with diverse backgrounds (architecture, interior design, building operations) We deploy on AWS Lambda to ensure high availability and high elasticity.
Layout & Furnishing AI
In today’s world, where we work is so much more than a set of desks, a sterile conference room and a sad kitchen. Designing dynamic, engaging, and culturally diverse workplaces is a complex and fascinating discipline that uses design, psychology, storytelling, human behaviour and so much more to create a space that feels exactly like it should.
We love that today’s workplaces are designed differently; they’re meant to be hubs of creation that are inspiring, interesting, forward thinking, usable and meet health and safety regulations, permitting, codes, and team-specific needs. It’s a big puzzle that gets put together room by room to make the whole.
As with anything that works at scale, we knew even workplace design could be streamlined and made more efficient. After studying the process of design, we discovered that we absolutely cannot automate the artistry and skill of interior designers and the teams that make magic happen, but we can automate some pieces of the puzzle. When it comes to design at this scale, there are a lot of repetitive actions that architects and designers shouldn’t have to do again and again. By automating these processes, we can save them significant time on the things that they don’t need to be spending their time on, giving them more space to do what they do best.
We’ve found that our Layout Automation works best with large spaces that have hundreds of equally designed aspects; from multiple conference rooms to seating areas, corridors, offices, and hallways, we’re able to automate aspects of design that don’t need an artistic eye.
Even better, and more practically, we’ve found that our automation tools make the turnaround time for a schematic first proposal good enough, accurate enough, and quick enough that you are able to provide the client with a fair estimate for your services.
How it works
In terms of process, we first design the zones — which includes rooms, corridors, offices, etc. Next, based on the function of the room, where the windows are, the entrance door, and similar factors, we utilise the algorithm and specific parameters to provide a design solution for the room. For this phase, you’re creating a simple proposal that meets basic workplace needs while utilising a basic set of furniture to design the space.
The algorithms used to furnish each of the specific usage type spaces all use the same underlying algorithm, which is a backtracking-type search. These algorithms are custom rule based for optimisation methods. In terms of deployment, the services are each packaged as a docker container and deployed on AWS ECS through AWS CodePipleline.
The system is a combination of search and optimisation algorithms and computational geometry which is developed as a data-driven approach with input from people with diverse backgrounds (architecture, interior design, building operations) We deploy on AWS Lambda to ensure high availability and high elasticity.
For the furniture group generation, we used API methods whose parameters are the group type, the bounding box, and the shape of the furniture pieces to choose from. Separately, there is a desk grid generation method for filling a rectangular space with desks, considering several parameters like width of the corridor, minimum space needed for a chair, and similar.
In development is a backend service that divides an office floor into particular usage zones and its code and API method are part of the Auto-Furnishing app. In addition, the recogniser app directory contains the code for the entry point and the post processing phase of a floor plan parsing services. This generates a 3D scene structure from a floor plan, necessitating a floor plan segmentation service.