Ship attitude prediction based on Input Delay Neural Network and measurements of gyroscopes

Due to the uncertainty and random nature of ocean waves, the accurate prediction of ship attitude is hard to be achieved, especially in high sea states. A ship attitude prediction method using Input Delay Neural Network (IDNN) is proposed in this paper. One of the advantages of this method is that it takes the measurements of Microelectromechanical Systems (MEMS) gyrosocpes, besides ship Euler angles, as the inputs of IDNN, which can greatly increase the prediction precision of ship attitude with little increase in system cost. The effectiveness of proposed method is validated through a data set sampled in a ship simulation hardware system. Moreover, the factors that affect the prediction performance are also explored through a set of experiments. The prediction method proposed can achieve high precision, that is, the root-mean-square prediction errors for roll, pitch and yaw, are 0.26 deg, 0.12 deg and 0.26 deg, respectively, when the prediction time is 2 sec. This precision is high enough for most attitude stabilization control systems.