Speaker Affiliation: Naval Undersea Warfare Center, Newport RI
Date: Wednesday April 18th, 2007
Time: Noon - 1:00 PM
Location: Room 5-314
Abstract:
Three trends are changing how unmanned underwater and surface vehicles are viewed and used by the science and military communities. First, the improvement in the cost/performance ratio means these systems are no longer exclusive to larger organizations. Second, the vehicles themselves are smaller, easier to use and deploying them no longer requires access to an expensive research vessel. The third trend is that acoustic communication in the sub-surface domain opens the door for collaboration between vehicles to observe larger areas in less time, and to use multiple vehicles to sense phenomena not easily sensed with a single vehicle. These trends present a research challenge in the autonomy algorithms needed to reach the potential of unmanned marine systems. The challenge concerns the algorithms themselves, which need to accommodate the collaborative, adaptive, long-term missions of ocean observation. It also concerns the nature in which autonomy is developed across the rapidly growing and distributed science community putting these systems to work.
In this talk we present autonomy methods developed for and used on unmanned marine vehicles based on a novel mathematical programming model for multi-objective optimization. Multi-objective problems arise when a single vehicle is balancing competing sensing, safety and operation objectives, and also occur in collaborative vehicles reconciling individual and group objectives. The interval programming
(IvP) model will be presented which uses piecewise linear objective functions to approximate individual vehicle autonomy objectives. A branch-and-bound solution method, free of function form assumptions, is discussed that exploits the piecewise linear constructs in its pruning algorithms to ensure globally optimal point decisions modulo the error introduced by the piecewise linear approximations. Results from in-water experiments controlling groups of autonomous surface craft and autonomous underwater vehicles will be presented.
Biography: Michael Benjamin received the BS and MS degrees in computer science and cognitive science from Rensselaer, and the MS and PhD degrees in computer science from Brown University. He was a post-doctoral fellow in the Ocean Engineering Department at MIT and is currently a visiting scientist in Mechanical Engineering at MIT. He is a research scientist for the Naval Undersea Warfare Center in Newport Rhode Island and also currently for the Office of Naval Research. His research interests include autonomy algorithms and software for unmanned underwater and surface vehicles and multi-objective optimization methods.
