Use of deep learning in autonomous vessels

Primjena tehnologije dubokog učenja u navigaciji i upravljanju pokazuje obećavajuće rezultate ne samo kod cestovnih i zračnih već i kod morskih vozila. Iako prekomorska plovidba postavlja jedinstvene zahtjeve sigurnosti prometa i robe postoji velika potreba za optimizacijom procesa koji se odvijaju na brodu uz smanjenje ljudskih resursa. S obzirom da ljudska pogreška pri odlučivanju snosi veliki dio uzroka nesreća i kvarova, područje razvijanja pametnih tehnologija koje bi zamijenile ljudski faktor u ovu svrhu predstavlja brzorastuću disciplinu. Naravno, ovaj pristup nosi svoje mane i prednosti koje su detaljno iznesene u radu. Cilj rada je pregled aktualnih metoda dubokog učenja te opis autonomnih vozila poznatih u današnjem pomorstvu.The application of deep learning technology in navigation and control shows promising results not only in road and air vehicles but also in sea vehicles. Although overseas travel sets unique requirements for the safety of traffic and goods, there is a great need to optimize the processes that take place onboard while reducing human resources. Given that human error in decision-making bears much of the cause of accidents and breakdowns, the area of developing smart technologies that would replace the human factor for this purpose is a rapidly growing discipline. Of course, this approach carries its disadvantages and advantages which are detailed in the paper. The aim of this paper is to review current methods of deep learning and a description of autonomous vehicles known in today's maritime industry