A prediction of ship motions with a Support Vector Machine

The Landing Periods Designator (LPD) is a system that aids the helicopter pilot during the landing operation on a ship, giving hints of the landing safety. At present, empirical indexes are used for this task, giving indications of the motion of the ship. This work faces the problem in another way, by computing ship motion predictions for a period of few seconds by using past motion information. This black-box approach is based on the Support Vector Machine (SVM) algorithm used as a time series predictor. The purpose of the SVM is to learn the signal dynamic by observing historical examples and then generate a prediction of the near-future dynamics. Experimental results performed both on simulations and real-world data show that our black box approach is comparable with more complex and special-purpose systems.