Machine learning of neural network classifier for intelligent autopilot on fuzzy logic

Sedova N. A., Bazhenov R. I., Dorofeev A. S., Glushkov S. V.

The article reflects the results of the development of an intelligent autopilot model, where a neural network classifier was used to adapt the regulator, which selects the parameters of such a regulator in accordance with neural network ships models obtained from their trajectories. The article reflects the results of the development of an intelligent autopilot model, where a neural network classifier was used to adapt the regulator, which selects the parameters of such a regulator in accordance with neural network ships models obtained from their trajectories. Using a certified signal simulator for autopilots, ships movement trajectories for different types of vessels under different sailing conditions were obtained. The optimization selection of the type, architecture and learning algorithm of the neural network for each trajectory was carried out. A spectral analysis showing the change in the spectral characteristics of ship models when changing weather conditions was done. The knowledge base of the neural network classifier of 24 neural network models of sea ships trajectories was formed, and usability of proposed approach was shown.

Keywords:navigation safety, intelligent autopilot, machine learning methods, classifier, fuzzy logic controller, neural network.