Submarine Maneuvers Prediction using Recursive Neural Networks

Recursive neural networks (RNNs) are a technique for developing time-dependent, nonlinear equation systems. In this paper, we applied RNN to simulate the maneuvers of submarine. The forces and moments acting on the body of submarine are functions of the motion state variables. Component force modules is developed to calculate five component forces as inputs to the recursive neural networks. These forces are related to the input control variables such as rudder angle, propeller revolution and the output state variables are the time histories of the motion velocities. These output data can be integrated to recover the trajectory and attitude, and differentiated to determine the acceleration acting on the submarine. The outputs of longitudinal velocity, lateral velocity and yaw rate are feed back to the input layer of the network beside the above forces. In this study, an existing submarine maneuvering simulation program which has been developed basing on US Navy Hydrodynamic Technology Centre (US NHTC) model is used for generating all the sample data of maneuvering for training and validation RNN. The results indicate that the RNN simulations provide fast and accurate predictions for submarine maneuvers used to develop the simulations as well as for validation maneuvers