Research on Ship Motion Prediction Algorithm Based on Dual-Pass Long Short-Term Memory Neural Network

Xiong Hu,Boyi Zhang,Gang Tang

doi:10.1109/access.2021.3055253

Xiong Hu, Boyi Zhang + Show 1 more

Open Access

https://doi.org/10.1109/access.2021.3055253

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2021
Citations: 32	License type: CC BY 4.0

Affiliation: Shanghai Maritime University

Abstract

When forecasting ship movements, the random errors of the inertial navigation system (INS) seriously affect the accuracy of general prediction methods. In actual measurement, the main causes of the random errors are electrostatic bias and micro-electric disturbance. In response to this problem, a novel type of dual-pass Long Short-Term Memory (LSTM) neural network architecture is developed, on the basis of regular LSTM neural network. In the designed dual-pass LSTM neural network, the random drift and the noise residual of the INS are regarded as a autoregressive moving average (ARMA) and generalized autoregressive conditional heteroskedasticity (GARCH) model. Through dual-pass layers, the prediction of drift and the correction of residual errors are realized respectively in the same time. The simulation of ship heave motion was carried out on the ship motion simulation platform, and the real-time datas which are measured by the INS are inputted to the trained dual-pass LSTM netural network. The experiment proved that, when training the same source datas offline, the average Root Mean Squared Error (RMSE) percentage of conventional LSTM network was 3.94%, but when training different source datas or training online, the prediction accuracy obvious decline. In contrast, the average RMSE percentage of the dual-pass LSTM neural network was 1.05% when training offline and 1.12% when training online. Compared with conventional LSTM networks, the dual-pass LSTM network is more targeted and has better adaptability in the field of ship-motion prediction, and this network restores the motion prediction to the actual trajectory of a ship more accurately.

Highlights

Nowadays, Inertial navigation system (INS) has high application value in the field of ship-motion prediction, owing to its significant superiorities like low depend on external information and no energy radiation outside [1]–[3]
2) On the basis of the original Long Short-Term Memory (LSTM) neural network, a transfer layer for residual calculation is added, and a network connection for residual correction is designed with reference to the generalized autoregressive conditional heteroskedasticity (GARCH) model, so that the improved dual-pass LSTM neural network can predict drift errors while simultaneously Correction and compensation of noise residuals
LSTM NEURAL NETWORK APPROXIMATION In Eq(2), the autoregressive moving average (ARMA) model constructs a type of time series predictor, which can evaluate and predict future data through the past time series and error series

Summary

INTRODUCTION

Inertial navigation system (INS) has high application value in the field of ship-motion prediction, owing to its significant superiorities like low depend on external information and no energy radiation outside [1]–[3]. The AI methods mainly include support vector machine modeling methods [18], neural networks modeling methods [19], [20], etc, which regard the processing of modeling as a problem of sequence prediction These methods aim to model and compensate the random drift as accurately as possible in offline conditions, there into the random drift is obtained after a preprocessing of removing white noise from the stochastic errors. According to the INS drift error and noise residual length and timing characteristics, a dual-transfer layer architecture LSTM neural network is designed. 2) On the basis of the original LSTM neural network, a transfer layer for residual calculation is added, and a network connection for residual correction is designed with reference to the GARCH model, so that the improved dual-pass LSTM neural network can predict drift errors while simultaneously Correction and compensation of noise residuals. 3) Experiments were conducted to verify the effectiveness and superiority of the improved dual-transfer layer LSTM neural network in predicting the random error of the INS system

ARMA MODEL AND LSTM NEURAL NETWORK APPROXIMATION

GARCH MODEL

CONSTRUCTION OF DUAL-PASS LSTM NETWORK

TRAINING PROCESS

RESULTS AND DISCUSSION

OFFLINE TRAINING