SIMULATION OF A NON-AXISYMMETRIC UNDERSEA VEHICLE USING A RECURSIVE NEURAL NETWORK

摘要

Turning maneuvers of a non-axisymmetric undersea vehicle have been conducted by Northrop Grumman Shipbuilding, and experimental data has been acquired from their free-running Nnemol model. This paper will focus on the application of a nonlinear time domain technique, based on a fast recursive neural network (RNN) approach, to simulate the six degree-of-freedom (6-dof) motion of the vehicle. Graphs are presented comparing the predicted motions with the experimental measurements, and error measures are used to quantify the results. The predictions clearly capture the details of the maneuvers and demonstrate a faster-than-real-time simulation capability. The simulation will be coupled with a controller to allow advanced, predictive control strategies to be explored.