Automatic Berthing System Using Artificial Neural Network

点击翻译按钮获取中文摘要

Articitial neural network (ANN) is applied for automatic berthing system for ships.1) It was verified that ANN is valid for berthing control for a normal shipswith one propeller and one rudder by simulation.2) It was also applied for a ship with Vec Twin rudder system (Vec Twin ship), which has one propeller and two rudders.・Mathematical model for a Vec Twin ship was developed and verified with model experiments.・Three-Layrs ANN was configured for a Vec Twin ship.・Simulation program for manual steering of a Vec Twin ship was developed.・Simulation program was modified for easy handling and preparing the teaching data.・After learning done by ANN,three-layr ANN was found to be satisfactory in the steering of a Vec Twin ship.3) Model experiments were planned to verify the simulation.・Several methods to obtain ship position were investigated.・The accuracy of GPS was checked, using RTK-OTF (Real-Time Kinematic On the Fly) GPS system. The accuracy was found to be enough for berthing manoeuvres. However, it is still very expensive to use it now.・Investigation and research on image processing position measuring system should be done further.・Ultra-sonic position measuring system is hard for the purpose of automatic berthing.・In the future, the model experiments using ANN are planned based on these investigations.