Machine Learning: Neural Network Models of Sea-Craft Paths

Abstract

AbstractIn this paper, the authors suggest practical techniques and machine learning algorithms to develop an intelligent autopilot steering control system. The system consists of a module with the neural network models of sea-craft movement patterns, a module with a neural network classifier, and a module containing an optimizer that calculates the best fuzzy controller operating. The paper presents the findings of reviewing neural network paths for six types of marine vessels. It emphasizes that it is necessary to develop an original neural network model on different maritime navigation conditions. Therefore, one requires a neural network classifier that can select an optimal neural network model of a marine vessel movement patterns under certain waterway navigation conditions. The authors figure out the neural network models of sea vessel movement patterns using a certified IC-2005 signal simulator for an autopilot. The researchers apply machine learning algorithms to generate parameters of neural network models and classify ship movement patterns. The authors also carry out a computer simulation of sea vessels with different speed and on various navigation conditions to test if it is easy to use the neural network classifier reported. The paper covers the correspondence of the mean square root error of learning to navigation conditions for various sea vessels, including a trawler with a slow response.KeywordsMachine learningNeural networkIntelligent systemMarine vessels