Dr. Jerome Le Ny
University of Pennsylvania
Thursday, March 3
One of the great challenges for robotic networks is to perform collaborative missions in unknown environments. We consider deployment problems where a mobile robotic network must optimize its configuration in a distributed way in order to minimize a steady-state cost function that depends on the spatial distribution of certain probabilistic events of interest. Application include source seeking, coverage control, spatial partitioning, and dynamic vehicle routing problems. Contrary to most approaches that depend on a precise knowledge of the environment, we assume here that the event distribution is a priori unknown and can only be progressively inferred from the observation of the actual event occurrences. We present a general framework to solve such problems based on stochastic gradient algorithms and stochastic approximations. The stochastic gradient view simplifies and generalizes previously proposed solutions, and is applicable to new complex scenarios, for example adaptive coverage involving heterogeneous agents. Finally, our algorithms often take the form of simple distributed rules that could be implemented on resource-limited platforms.
Jerome Le Ny received his Bachelors degree from the Ecole Polytechnique, France in 2001, M.Sc. in Electrical Engineering from the University of Michigan-Ann Arbor, in 2003, and Ph.D. in Aeronautics and Astronautics from MIT in 2008. From December 2003 to August 2004, he was an embedded software engineer with Robert Bosch GmbH. He is currently a Postdoctoral Researcher with the GRASP Laboratory at the University of Pennsylvania. His research interests include robust and stochastic control with applications to autonomous and embedded systems, air transportation, and more generally computational methods supporting the design and verification of complex systems.