Fig. 1 [8]    Fig.2 [8]
 Fig.3 [8]  Fig. 4 [8]
 
Fig. 5 [8]                                          Fig. 6 [8]
 
 

 

 The LaHave House Project explores the creation of an automated architectural design service based on an industrial design approach to architectural practice in which Architects design families of similarly structured objects, rather than individual ones, thereby lowering design costs. At the heart of the project is the use of generative grammars to build design libraries. All of the images on this page are of houses resulting from the LaHave House Grammar.[7]
Currently in North America architects are involve in the design of only about 5% percent of the total new house market. Whereas custom architectural design will always have a premier role to play, we believe that an industrial design approach to architecture can bring much of the design quality and variety of custom design to the other 95% of the market, at an affordable price. 
Generating House Designs:
1- LaHave House Grammar
In this section, an example of the LaHave house desgin will be described using a single shape grammar to drive the house design. One feature of the grammar is its tendency to produce dense cores for servies and open spaces for living in. The grammar is constructed in terms of a set of five elementary components: Rooms, Tartans, Machines, Bays, and Totems. The Rooms are the principle places for human action, the tartans provide space for circulations, the machines (bathrooms, kitchens, laundry, entry etc.) are the dense service spaces, and the totems (staircases, cabinetry, etc) provide focus for the rooms. The grammar makes extensive use of the idea of served vs. servant spaces. For example, a dinning room may be served by a kitchen, which is in turn served by a pantry. The grammar of this example has a longitudial axis of growth, with a dense machine zone flanking a body zone, which is in turn surrounded by a bay zone (see figure 7, 8). [9]
            Fig. 7 - Example of a floor plan.[7]                 Fig. 8 - Space zoning. [7]
2- Phases of Generations
A hierarchical set of shape grammars are used to generate both plan and section schema. A set of backward rules are then used to identify within plan possible groupings of individual spaces into abstract functional zones. Given compatible plan and section, a form can now be created. This entails creating floor plans, dimensioning these plans, creating walls to define the interior spaces, adding roofs and generally computing all of the 3D information required to define a complete form. Finally, forms are transformed into completed house designs by a rule based system that first assigns to each space unit within the form a function (i.e kitchen, bedroom, entry, etc.)(see figure 9).[9]
Fig. 9 - Phases of Generation. [7]                    Fig. 10 - Example of an Invalid Section. [7]
After generating the house schema, comes the validation of the desgin. Figure 10 presents an example of two invalid features. the left hand secondary form is too large for the primary form it connects to, and the roof of the primary form is too shadow to support the story [7]:
the rules which prevent these features form occuring are given below [7].

invalid_section
    if roof is not flat and 
        stories in body < stories in bump.

invalid_section
    if distance from roof-peek to next floor < 8'.
 

Fig.11 [8]
The software has been tested in a trial to build in a resort community ninety minutes outside Seoul, South Korea.[8]
 
Home • AI • Structure • Case Study • Conclusion • Reference
site implementation as a project for Philosophy of Structure