Interactive Evolution by Duplication and Diversification of L-systems

Thomas Burt

Abstract

Evolutionary processes are responsible for the wide variety of organic forms seen in the natural world. Digital simulations inspired by these processes can be used to create diverse models for computer imagery purposes. In the case of interactive evolutionary computation (IEC), a modeler affects the resulting forms by guiding evolutionary processes. This guidance can be achieved using very simple interfaces, thus putting IEC in the hands of a broad class of users.

In this work we focus on the application of IEC to plant modeling. The individual plants are generated using L-systems. Evolution of L-systems is particularly appealing due to analogies that can be drawn between genetic code and L-system productions. Several attempts to evolve L-system-based models have been devised in the past by different authors using the standard operations of genetic algorithms --- mutations and crossover --- to modify the productions. We have improved on these results by considering a different set of operations --- duplication and diversification --- as the basis of L-system evolution instead. These operations reflect the widely accepted theory of evolution of genetic material, according to which gene duplication is the first step in smooth transitions from less complex to more complex phenotypes via gradual diversification of redundant genetic material. This thesis investiages how this general principle can be adapted to the evolution of L-systems, presents an interactive modeling system based on the resulting theory, and illustrates the results using numerous visual examples.

Reference

Thomas Burt. Interactive Evolution by Duplication and Diversification of L-systems. M.Sc. thesis, University of Calgary, July 2013.

Download PDF here (32 Mb), or download a lower quality version (7 Mb).