Optimal Control of a Mobile Sensor Network for Track Coverage

Kelli Baumgartner
Special Instructions: 
Lunch and beverages will be served
Friday, October 19, 2007 - 12:00pm
Hudson Hall Room 216
Seminar Contact(s): 
Elizabeth Irish eri@duke.edu or Justin Jaworski jwj@duke.edu
Semester & Year: 
Fall 2007
The problem of optimizing the configuration of a moving sensor network deployed to detect moving targets is formulated using optimal control theory. A Lagrangian cost function is formulated through a computational geometry approach to measure the space of line transversals for k of the n sensors by formulating an integral function of the sensors locations, where k is the number of required detections. Then, included in the cost function integrand is a penalty on energy such that the multiobjective cost function with competing objectives is optimized subject to the sensors dynamics expressed by a state-space model. The method is demonstrated for a surveillance application that involves group of underwater gliders, each with an onboard acoustic omnidirectional sensor, deployed on the ocean’s surface to detect underwater targets within a specified region of interest and over a desired period of time. It is shown that the nonlinear, time-variant ocean current velocity field can be modeled by a neural network trained by CODAR measurements. A solution is presented for both the non-maneuverable (sensors placed on sonobuoys) and mobile (sensors onboard gliders) sensor networks that can be placed anywhere within the region of interest and move subject to the ocean’s current and possibly control.