Despite the rising popularity and utilization of intelligent systems, much of the built environment, especially architecture, remains prescriptive or responsive in nature. Kinetic facades, especially, still rely on the analysis of historic or approximated data to generate a solution through the utilization of a multi-objective optimization (MOO) algorithm. This approach lacks the ability to adapt to the changing forces (i.e. site specific micro-climates or changing occupants) to which buildings are subjected for two reasons: 1) MOO is computationally expensive due to the immense solution space and 2) it can only solve for known objectives. This lack in ability for facades to adapt to changing conditions or be designed using actual site data has been one of the hindrances on the growth of the industry.

Kinetic facades should instead be developed as an integrated portion of an intelligent system that is able to unify environmental data and user input in real-time to create an interior environment that is comfortable, energy efficient, and able to adapt to any future changes. Inputs such as space volume and user preferences, then, must be assumed to be unknown, putting MOO at a disadvantage in intelligent systems. As such, I have developed a control system architecture for an intelligent facade that utilizes a neural network algorithm (a form of machine learning) to address the need of adaptation in kinetic facades. To test my method, I utilized Rhino 5, Grasshopper, and Python with a simulated dataset as a case study.