PREDSTAVITEV MODELOV UMETNE INTELIGENCE ZA UPORABO AVTONOMNEGA MANEVRIRANJA PLOVIL

Cizelj, Domen (author), Krmac, Evelin (mentor)

V diplomskem delu je na splošno predstavljen sistem za avtonomno izogibanje trčenju in algoritmi, s katerimi je tak sistem lahko zasnovan. Predstavitev se začne z opredelitvijo obravnavanega področja, nameni, cilji, predpostavkami in opisom raziskovalne metodologije. Sledi splošna predstavitev področja avtonomnih plovil in razlogov za njegovo rast in razvijanje. V nadaljevanju so podrobneje predstavljene stopnje avtonomnosti in delovanje avtonomnih plovil ter sistem za avtonomno izogibanje trčenja. Predstavljeni so tudi pomembni senzorji, potrebni za delovanje takega sistema. Sledi osnovna razlaga algoritma in algoritma strojnega učenja (algoritem SU) (angl. Machine Learning algorithm). Na koncu tega poglavja je podana tudi splošna sestava algoritma za avtonomno izogibanje trčenja. Sledi predstavitev posameznih algoritmov, uporabljenih na področju razvoja avtonomne plovbe in avtonomne vožnje, med katerimi so Genetski algoritem, algoritem »K-tega najbližjega soseda« (algoritem KNS) (angl. K-nearest neighbor algorithm), »Metoda podpornih vektorjev« (MPV) (angl. Support Vector Machines), algoritem »potencialnih umetnih polj« (algoritem PUP) (angl. Artificial potencial field algorithm) in pa globoke nevronske mreže (GNM) (angl. Deep Neural Networks). Pri vsakem algoritmu je podana tehnična razlaga, preprosta razlaga in predstavitev možne uporabe algoritma na področju avtonomnih plovil. V zaključnem delu je podana primerjava in analiza predstavljenih algoritmov.

• Language: Slovenian
• Keywords: Avtonomna plovila, algoritmi, algoritmi strojnega učenja, umetna inteligenca, sistem za avtonomno izogibanje trčenja na morju.
• Work type: Bachelor thesis/paper
• Organization: FPP - Faculty of Maritime Studies and Transport
• Year: 2023
• PID: 20.500.12556/RUL-151662
• Publication date in RUL: 14.10.2023

Secondary language

• Language: English
• Title: PRESENTATION OF ARTIFICIAL INTELLIGENCE MODELS FOR THE APPLICATION OF AUTONOMOUS MANEUVERING OF VESSELS

Abstract

The thesis generally presents a system for autonomous collision avoidance and the algorithms that can be used to design such a system. The presentation begins with a definition of the area under consideration, purposes, goals, assumptions and a description of the research methodology. The following is a general presentation of the field of autonomous vessels and the reasons for its growth and development. In the following, the levels of autonomy and the operation of autonomous vessels, as well as the system for autonomous collision avoidance, are presented in more detail. Important sensors necessary for the operation of such a system are also presented. The following is a basic explanation of the algorithm and the machine learning algorithm. At the end of this chapter, the general structure of the algorithm for autonomous collision avoidance is also given. The following is a presentation of individual algorithms used in the development of autonomous navigation and autonomous driving, among which are the Genetic Algorithm, K-Nearest Neighbor algorithm, Support Vector Machines, Artificial Potential Fields algorithm and Deep Neural Networks. For each algorithm, a technical explanation, a simple explanation and a presentation of the possible use of the algorithm in the field of autonomous vessels are given. The final part provides a comparison and analysis of the presented algorithms.

• Keywords: Autonomous vessels, algorithms, machine learning algorithms, artificial intelligence, autonomous collision avoidance system.