Home | Meeting Announcement | Meeting Agenda | Call for Papers | Meeting Registration | Hotel Information | Travel Information | Attendees | Abstracts
2002 LAPCOD Meeting
Previous Abstract | Back to Abstracts Page | Next Abstract
Efficient Navigation and Adaptive Sampling Strategies for Fleets of Autonomous Underwater Gliders
E. Fiorelli(a), S. Shadden(b), N.E. Leonard(a), J.E. Marsden(b)
(a)Princeton University, (b)California Institute of Technology
eddie@princeton.edu(Abstract received 10/31/2002 for session A)
ABSTRACT
An underwater glider is a fixed-wing, buoyancy-driven autonomous underwater vehicle with the ability to redistribute its internal mass to effect net changes in attitude. In addition, fixed wings provide lift to induce forward motion for non-zero angles of attack. By design, underwater gliders are reliable and efficient--making them well-suited for long duration ocean sampling, especially in the context of multi-vehicle, mobile, adaptive sampling networks.
In this talk we describe how fleets of coordinated underwater gliders can act as ocean sensors for adaptive sampling. Ocean currents undoubtedly affect glider motion; hence it is important to utilize the natural dynamics of the flow when developing efficient navigation schemes. We propose a strategy that uses Direct Lyapunov Exponent (DLE) contours to lead the gliders toward areas of sampling interest, e.g. fronts. Repelling material lines obtained from DLE maps computed forward in time are used as navigation channels. The gliders track these repelling material lines to reach fronts that coincide with attracting material lines, which are obtained from DLE maps computed backward in time. We present control methods that enable groups of gliders to climb gradients of DLE fields as one mean of tracking the navigation channels.
The control laws that we propose to coordinate the gliders derive from artificial potentials. The idea is to impose simple feedback rules at the individual glider level that enable more complex group-level behaviors such as formation maintenance, group translation, rotation and expansion in response to measured data. These types of behaviors can then be put together to perform tasks such as gradient climbing or sampling in and around a front or feature of interest. We illustrate the latter idea by considering an example whereby we induce a formation to roll along a sensed front (e.g. temperature). The net group motion is along the front but each vehicle visits the front boundaries repeatedly, thus sampling both the interior and across the boundaries.
Previous Abstract | Back to Abstracts Page | Next Abstract
2002 LAPCOD Meeting, Key Largo, Florida, December 12-16, 2000