Research On Ship Collision Avoidance Based On Deep Learning

The current shipping industry is featured with two characteristics.The ship general control and management has been increasingly difficult due to the enlarged ship size and the advancement of technology of modern ship.Moreover,the sailing time of oceangoing ships at deep sea is increased since ship’s voyages and the stopping points rises because of the change of the global trade pattern.In this case,the incidence of ship collision is on the rise.Taken the ship collision avoidance as a research object,the method of collision risk index(CRI)calculation and the scheme of the ship assistant autonomous collision avoidance system are investigated in this thesis.The main contents are listed as follows:1.An improved CRI formula is presented after the traditional way is analyzed.Firstly,the whole process of ship collision and the principle of collision avoidance is analyzed including various encounter scenes and each phase of collision process.Secondly,the application of the major ways of collision risk assessment has been studied.Then a new thought of calculating ship collision CRI is proposed after covering the deficiency of major ways to calculate ship CRI.2.The calculation method of ship CRI is proposed based on Deep Neural Network(DNN).Firstly,the application and deficiency of traditional Artificial Neural Network(ANN)on CRI calculation is analyzed.Then a method of calculating collision risk index based on DNN is presented ranging from DNN model establishing to model verifying.Finally,the DNN training model is tested with the ship navigation simulator.3.A complete scheme of ship assistant autonomous collision avoidance system is presented.Considering the newly developed shipping technology,a complete scheme of ship’s autonomous collision avoidance is proposed covering information awareness,knowledge base,encounter scenes analysis,collision avoidance index assessment,decision-making systems,emergency systems,and so on.