Computer generated architecture through nature: using swarm intelligence and evolutionary computing to generate 3D models

Jason Palevsky, The College of Wooster

Abstract

This paper examines the nature inspired machine intelligence methods of swarm intelligence and evolutionary computing. Swarm intelligence algorithms utilize stigmergy, indirect communication through environmental changes, to create emergent, complex, cooperative behavior amongst primitive, ``dumb'' actors. Some examples of such algorithms are Ant Colony Optimization, based on ant foraging behavior and the use of trail pheromones, Firefly Algorithm, based on the bioluminescence fireflies use to attract other fireflies and the concept of light dispersal with distance, and Wasp Nest Building Algorithms, based on the cooperative nest building of many wasp species. Contrary to popular belief, when building nests experimental evidence has indicated that it is not due to an internal blueprint but rather reactions to environmental states that drives the actions of each wasp -- in other words, wasps employ stigmergy to build their nests. The nest building algorithm utilizes a population of agents moving about in a 3-dimensional (3D) environment and whom, given a set of ``building rules'', place building blocks in the environment in order to construct an architectural form. Evolutionary computing employs methods from Darwinian evolution and genetic theory to produce a solution to a problem through the creation of successive populations of possible candidate solutions. The problem controls the evolutionary process through a fitness function that evaluates the quality candidates' solutions and drives selection and reproduction opportunities. This paper examines the implementation and applications of these nature-inspired swarm intelligence algorithms in conjunction with evolutionary computing and the potential to implement an evolutionary wasp nest building algorithm that evolves the ability to generate a desired architectural form.