An introduction to Catalyst, our automatic layout designer
As an industry-leader in space planning and visualisation one of our key products is Catalyst, an automatic bathroom layout design tool. Using machine learning it guides a customer to plan their perfect room helping reduce the friction that occurs during the design process. Just like a real-life designer, we get to know the customer and combine their room dimensions, budget, tastes and needs to recommend specific products, the layout of which is automatically optimised to meet regulations and good layout guidelines.
This blog post will outline the rules and regulations we consider when designing a customer’s bathroom.
So what does Catalyst actually do?
Catalyst uses design rules to automatically generate a layout for a given room. To do so it requires:
A description of the room. This includes the placement of the walls, doors and windows.
A description of the possible products that could be placed into the room. This includes the dimensions of the product and whether the object should be mounted on one wall or two.
An algorithm is then applied to find the combination and layout of products that best fits our set of design rules for this specific room.
The below gif demonstrates the algorithm running on two different rooms. The difference between the two rooms is the placement of the door and the window. The set of possible products is the same in both cases. We can see that in one room a bath is preferred, however in the other a shower is preferred. This shows the power of the algorithm to choose the products based upon the layout of the room.
Once the core products (toilet, basin, etc.) are placed we add finishing touches such as a tooth brush holder to give a more ‘lived-in’ feel to the room.
But what design rules are included?
The design rules we include are based on legal regulations or have been specified by bathroom design experts. A design rule may be a hard constraint, i.e., the layout is only valid if this is true, or a soft constraint, i.e., the layout is preferred if this is true but it is not completely necessary.
One of the most important considerations when designing a bathroom is that each product has adequate space to be safely used as intended. The minimum amount of space required for each type of product is outlined in design regulations and we use a hard constraint to ensure this space is not violated. However most designers agree that the space defined by the legal regulations is insufficient to comfortably use a product and recommend additional space. Subsequently, if possible we leave more space between each of the products or the walls of the room. The image to the left illustrates the required and recommended space for a toilet.
The size and shape of a room can often be used as a cue for the products that should be placed in the room. For example, a family bathroom may be large and contain a bath; whereas an en-suite bathroom would usually be smaller and contain a shower cubicle. We handle this by finding an equilibrium between two design rules:
We penalise layouts that do not include each of a specified set of product types.
We prefer layouts that have an unoccupied floor space percentage similar to that of a typical bathroom. As shown in the image to the right, the unoccupied floor space is the floor space where no products are placed.
Through these two design rules we are able to choose a combination of products suitable for the specified room, yet we must also consider the distribution of the products around the room. For example in a large room a customer is unlikely to want all the products located along one wall. More likely they want the products distributed around the room so they can maximise the use of the space. While the recommended space design rule partially handles this, we add another design rule to prefer layouts that disperse the products around the room.
Finally, we include a set of design rules specific to certain types of products.
A layout is preferred if the toilet is not in a hypothetical corridor in front of the door.
A layout is preferred if the shower is located in a corner of the room.
A layout is preferred if the toilet is placed on a wall that also has a window. At present the algorithm is not provided with the position of the soil-pipe in the room. As such, we assume that the soil-pipe would be on an external wall and use windows as a proxy for the external walls.
We have a set of rules that define the relationship between different types of products. A layout is preferred if this relationship is true. For example, we prefer layouts where a cabinet can be placed above the basin. This encourages layouts that do not place the basin under a window so a cabinet can be added. However if no other alternative exists then the basin will be placed under the window and no cabinet will be added to the layout.
Where do we go from here?
The development of the Catalyst automatic layout designer algorithm is an iterative process. As described earlier, we don’t currently include the position of the soil-pipe in the description of the room. Therefore we prefer room layouts where the toilet is mounted on a wall with a window in the assumption that this is an external wall. We are currently working on integrating a more precise soil-pipe location. This is an important change as moving the soil-pipe within a room can be a very costly procedure. Given the soil-pipe location we will lock the position of the toilet and the other products will be optimised given the toilet’s position as an additional design rule.
Similarly there may be situations where a customer wishes the position of some products to be fixed in the room. For example, they may not be replacing their bath so this should be locked in place. In this case the algorithm will fix the position of the existing products and optimise the other products in the room.
We are also performing preliminary work to apply Catalyst to kitchens. This a more complex problem with a much larger range of possible products and a larger set of design rules. We believe we have the in-house expertise to tackle this problem and will have this available as a product in the future.