UCL Discovery

Abstract

While is often assumed that adaptation to a complex set of phenomena requires a complex control mechanism, Adaptive Fa[CA]de suggests a simpler control mechanism in terms of independent units, yet more contextual to its environment. Rather than being a constantly moving structure which would waste energy and lead to potential breakdown, the façade is trained to anticipate its own future behaviour and therefore move less to adapt. To achieve the above, the inherent structural and performative characteristics of CA are used as means to obtain optimum light levels to the interior of the building. The façade is composed by a finite grid of panels that switch through a number of possible states translated in tilting angles. The hypothesis is that the facade can be trained by artificial Neural Networks and use the complexity of CA to take minimum analog input of the environment, translate it to digital and – through optimisation – back to analog in the form of different CA patterns. Due to CA’s emergent nature, the information is distributed throughout the grid of panels. Learning the irregularity of both the complex embedded structural attributes of CA and external light conditions, the system can then propagate successful patterns. Complexity here is used as an interface between different layers of information and suggests an alternative way to deal with the complexity found in most architectural configurations.