Echo State Network ship motion modeling prediction based on Kalman filter
Xiuyan Peng, Huiyuan Dong, Biao Zhang · 2017
According to the nonlinear characteristics of ship motion, Echo State Network (ESN) is proposed for modeling in ship motion. Meanwhile, ESN training algorithm exists deficiency at present. So it is Kalman filter algorithm that is applied to recursive training network output connected weight. Combining the phase-space reconstruction, prediction model of Echo State Network in ship motion is established. Testing simulation results show that the prediction model of Echo State Network based on Kalman filter algorithm which is proposed in this paper, improving the accuracy and length of forecast. A feasible method is provided for real time on-line modeling prediction in the ship motion.