An Ecological Approach to Generative Design

Abstract

In this paper we explore the use of an animat model to construct geometry. The model uses agents guided by direct (or active) visual perception of their environment to replicate human behaviour within a notional plan of an open space. The environment reacts to these agents by placing walls in order to affect their usage of the space, and thus the structure may be evolved to fit the social function of the agents within it. Here, we start with the most basic social function, to design a building that disperses agents programmed with an exploratory task across its floorplan by using an evolutionary algorithm. We investigate the effect of evolution on the generated configuration using space syntax tools. We show how the introduction of a simple rule, the desire to leave, can result in the evolution of commonly observed features — first a central axis and then a ‘foyer’.We discover that ‘intelligibility’ of the space, which might imply reduced cognitive load, may increase as the system is allowed to evolve. Finally, we consider the implications of the ecological approach for the design process.