(C303) <<< Previous | Back to Abstracts Page | Next >>> (C305)
The Inverse Lagrangian Prediction Problem
A.J. Mariano, T. Ozgokmen, T.M. Chin, A Griffa, E.H. Ryan
RSMAS, U. of Miami
amariano@rsmas.miami.edu(Abstract received 05/04/2005 for session D)
ABSTRACT
An inverse Lagrangian prediction problem is posed as follows: What is the initial location and deployment time so that the object will be found at a given location later in time? The "final" location/time of the object, such as a ship or drifting instrumented buoy, its drift characteristics, and noisy observation/simulation of the currents and winds are assumed to be known. Also, a multiple-object version of this problem would be: Can we design an optimal deployment strategy for a cluster of objects so that the objects end up in a specific configuration at some later time? A unique and/or numerically stable solution for such inversion problems is difficult to obtain computationally, given imperfection in the current and wind data, the chaotic nature of the forward prediction problem, convergent and divergent flows, deployment constraints, and bi-modal, down-wind drag coefficients. An "ensemble" of solutions describing a set of possible deployment locations/times for each final location might be the best one can achieve. A brute-force optimization technique that minimizes the mean-square distances between predicted final locations from a Monte-Carlo based simulation of trajectories and the desired final location is formulated and applied with good results to a practical problem in the Gulf of Mexico. Our optimization technique is applied to the array configuration problem, the central theme of ONR's Optimal Deployment of Drifting Acoustic Sensors (ODDAS) initiative.
(C303) <<< Previous | Back to Abstracts Page | Next >>> (C305)
2005 LAPCOD Meeting, Lerici, Italy, June 13-17, 2005