Grammatical inference methodology for control systems

摘要

Abstract:- Machine Learning is a computational methodology that provides automatic means of improving programmed tasks from experience. As a subfield of Machine Learning, Grammatical Inference (GI) attempts to learn structural models, such as grammars, from diverse data patterns, such as speech, artificial and natural languages, sequences provided by bioinformatics databases, amongst others. Here we are interested in identifying artificial languages from sets of positive and eventually negative samples of sentences. The present research intends to evaluate the effectiveness and usefulness of grammatical inference (GI) in control systems. The ultimate far-reaching goal addresses the issue of robots for self-assembly purposes. At least two benefits are to be drawn. First, on the epistemological level, it unifies two apparently distinct scientific communities, namely formal languages theory and robot control communities. Second, on the technological level, blending research from both fields results in the appearance of a richer community, as has been proven by the emergence of many multidisciplinary fields. Can we integrate diversified works dealing with robotic self-assembly while concentrating on grammars as an alternative control methodology? Our aim is to answer positively this central question. As far as this paper is concerned, we set out the broad methodological lines of the research while stressing the integration of these different approaches into one single unifying entity. Key-Words:- Machine learning, robot control, grammatical inference, graph grammar, formal languages for control, self-assembly, intelligent control, emergent control technologies. 1