1423 Powhatan St., Suite 1
Alexandria, Virginia 22314
Phone (703) 836-6727
Fax (703) 836-7491
Email: asnehq@navalengineers.org

Title: Neural-Network Technology Applied to Small-Boat Motion Prediction and Control

Author(s): Daniel A. Liut and Paul E. Jones, Leidos

Abstract: Launch and Recovery (L&R) operations can greatly benefit from the foreknowledge of vessel motions. Due to the large motions that can be induced by the wave field, motion predictions can be valuable information to aid the execution of maneuvers and the operation of onboard systems. A typical example is the execution of L&R operations involving a larger vessels and a small boat, due to the significant differences in their relative responses to the sea. Motion predictions of vessels can directly be used in the execution of a maneuver or be used to improve control systems, such as those utilizing rudders or stabilizing fins, in order to minimize undesired motions that could delay or impact L&R operations. Nonlinear effects, especially in larger sea states, can be dominant factors in smaller vessels. In addition to the large variability present in the excitation, as that associated to the different sea states, a small platform can exhibit its own variability due to changes in mass (e.g. from the addition or removal of cargo loads) and/or due to variations in weight distributions. The utilization of systems that can adapt to variations in environmental conditions and in the vessel’s dynamic and hydrodynamic characteristics are not only desirable but can become critical if such changes could alter the vehicle’s behavior noticeably in short periods of time. Due to their ability to capture nonlinear effects and to adapt quickly to changes via adequate training systems, neural networks appear as a promising technology that can be used for motion prediction and control of small (and larger) vessels, with a direct application to L&R operations. In the proposed presentation, a neural-network technology designed to meet the challenges described will be discussed. Its implementation, IMoPaC (Intelligent Motion Prediction and Control), has been used to predict ship motions and effect control. A small-boat application that utilizes such technology will be discussed with an emphasis on its controllability, mainly for course keeping and/or roll mitigation using only a rudder. IMoPaC has shown the ability to improve the performance that classical control systems can provide. With minimum tuning, it can “learn” its control operations and adapt to changes in the boat and the environment. Results from at-sea tests will be discussed, as well as results from simulations (used to help develop the technology) using the Large Amplitude Motion Prediction Program LAMP.