I. Introduction
Water transport has an important position and role in the national economy and logistics system[1], and is an important strategic resource for achieving sustainable economic and social development. In recent years, with the drastic increase in the volume of water traffic, the safe navigation and supervision of ships are facing a severe test[2], especially in the complex waters is more obvious. Therefore, it is of great significance to be able to timely and effectively predict the ship’s navigation trajectory in complex waters, which can help to reduce water traffic accidents, improve ship scheduling throughput and inland waterway capacity, and enhance the ship’s automation[3] and intelligent supervision ability, so as to realise the intelligent navigation of ships in the future.