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.