An Improved Real Time AR Method for the Surface Vessel Motion Prediction

It is significant and valuable to improve the time length and accuracy of ships' motion prediction for the efficiency, comfort and security of marine operation. Autoregressive time series analysis method (AR) is the mainstream currently and the effectiveness for the prediction of ships' motion attitudes has been fully validated. However, the algorithm fixes the order only once but forecasts the future data for multi-step, result in degradation of the step length and accuracy, especially when the ship sails in the bad sea condition. In order to solve this issue, this paper proposes a new autoregressive-multiple method (Arm), which can determine the orders and parameters of model in a real-time. The method is applied to forecast a ship's motion attitudes in eight different situations. The simulative results of autoregressive-multiple method show the validity and veracity.