Neural network application to ship position estimation

Abstract

The real time estimation of ship motion is considered in this paper. The problem is viewed as an estimatiod prediction problem for partially unknown systems. A neural estimator based on a dynamic recurrent neural network is considered. The model that describes the ship motion dynamics is presented, and the neural algorithm is tested and evaluated via extensive simulations. The results show that the new algorithm has excellent performance, and a significant saving in computational time is achieved.

Neural network estimators: application to ship position estimation

“…Motivated by the recent interest on neural networks application to state estimation problem [16]- [20] a dynamic neural estimator was used in [20] to provide estimates of the current and future ship position. The estimator was an input recurrent multilayer recurrent neural network, trained using back-propagation technique, to provide estimates of the ship position, Since in that analysis the comparative evaluation between neural and the statistical filters was the main objective, the dynamic model of the ship motion was completely known.…”