Recently, minimum time berthing problems have been numerically solved, however, a detailed ship motion model is necessary to obtain an optimal solution. To obtain the detailed model, detailed ship data by tank test when ship is designed, is needed. These detailed ship data are difficult to obtain a ship's captain or pilot. Therefore, it is difficult to obtain the solution of minimum time berthing problem for a ship's captain or pilot. In this research, the estimation of the ship motion model is tried from ship's principal particulars, which mariners can easily acquire. However, the model estimated from ship's principal particulars has possibility to include error margin. Therefore, in this research, it is tried to use the neural network controller to correct the minimum time control solution, which contains the model error. Using a small training ship, the validity of the estimated model and the neural network controller are examined by actual sea tests.
抄録全体を表示