on 29-04-2016
Friday, April 29, at 15h00.
Automatic methods are a promising approach to the design of robot swarms. In my seminar, I will present some of the results we have recently obtained at IRIDIA and I will sketch our ongoing and forthcoming investigations. Our research originated from the intuition that the reality gap problem bears resemblance to the generalization problem faced in machine learning. In particular, we devised na approach, called AutoMoDe, that is inspired by a strategy commonly adopted in supervised learning to deal with the so called bias/variance tradedoff: to obtain solutions that generalize well, an appropriate bias is injected in the design process. AutoMoDe produces robot control software by selecting, instantiating, and combining preexisting parametric modules — the injected biias. The resulting control software is a probabilistic finite state machine in which the topology, the transition rules, and the values of the parameters are obtained automatically via an optimization process that maximizes a task-specific objective function. I will present two instances of AutoMoDe, called Vanilla and Chocolate. They both design control software for a swarm of epuck robots. I will present also results obtained for several tasks including aggregation, foraging, área coverage, and sheltering. To conclude, I will sketch the core ideas that I will explore within DEMIURGE, a project funded by the European Commission via an ERC Consolidator grant that I was recently awarded.
BIO
Mauro Birattari received his master's degree in Electrical and Electronic Engineering from Politecnico di Milano, Italy, in 1997, and his doctoral degree in Information Technologies from the Faculty of Engineering of the Université Libre de Bruxelles, Belgium, in 2004. He is currently with IRIDIA, Universite Libre de Bruxelles as a Senior Research Associate of the Belgian Fonds de la Recherche Scientifique–FNRS. Dr. Birattari co-authored more than 100> peer- viewed scientific publications in the field of computational intelligence. His research interests focus on swarm intelligence, robotics, machine learning, and on the application of artificial intelligence techniques to the automatic design of algorithms. Dr. Birattari has been recently awarded an ERC Consolidator Grant by the European Research Council for the project "DEMIURGE: automatic design of robot swarmsâ€.