Splinter: A generic framework for evolving modular finite state machines

Evolutionary Programming (EP) has been used to solve a large variety of problems. This technique uses concepts of Darwin's theory to evolve finite state machines (FSMs). However, most works develop tailor-made EP frameworks to solve specific problems. These frameworks generally require significant modifications in their kernel to be adapted to other domains. To easy reuse and to allow modularity, modular FSMs were introduced. They are fundamental to get more generic EP frameworks. In this paper, we introduce the framework Splinter, capable of evolving modular FSMs. It can be easily configured to solve different problems. We illustrate this by presenting results from the use of Splinter for two problems: the artificial ant problem and the sequence of characters. The results validate the Splinter implementation and show that the modularity benefits do not decrease the performance.