The horizontal dynamics of deeply-towed underwater vehicle systems can be effectively modeled by nonlinear partial differential equations. However, high resolution numerical solutions are of limited use in controller design, where methods for systems of very high or infinite order are not well developed. This paper examines the possibility of finding low-order dynamic models in differential equation form, in order to present a more tractable control problem. A learning model method is used, and the identification process is novel in that a verified high-order model provides the primary data set. This approach allows a priori characterization of system responses in regimes or scenarios for which no experimental data exist. The performances of the reduced-order forms are verified through comparison of model output with actual sea data obtained during recent deepwater tests.


Towed underwater vehicles have many applications in science and industry, for shallow and deep water. When depths on the order of several thousand meters are involved, the inertial and drag forces on the cable may be very large. Therefore, accurate models of the cable and vehicle are important for tasks in which the vehicle position is controlled by maneuvering the ship. As an example case, we consider the horizontal control of MEDEA/JASON, a deeply-towed two-vehicle system developed at the Woods Hole Oceanographic Institution. As illustrated in Figure 1, the heavy MEDEA vehicle hangs from a taut cable terminating at the surface ship; JASON operates on a l00-meter tether from MEDEA. JASON is neutrally buoyant and carries thrusters, while MEDEA serves as an instrumented clump weight. This configuration decouples high frequency heave motions of the ship from JASON, and, at the same time, provides high cable tension to keep the horizontal offset small.

This content is only available via PDF.
You can access this article if you purchase or spend a download.