Abstract
Ship berthing is one of the most difficult ship operations. In the previous works, the artificial neural network (ANN) is verified to make it automatically. However, there are still many obstacles before it will be practically used. One of the problems is neither sea trial nor model ship experiment was done before. We have conducted model ship experiments based on ANN, and discuss for the availability in various conditions, especially for the method how to provide suitable teaching data. Combining model ship experiments and simulation, we have concluded that ANN can be practically applicable for automatic ship berthing. Secondary, we have discussed on the structure design of ANN for this purpose and propose a new ANN structure rationally.
