Research On Ship Route Planning Method Based On AIS Data

The maritime industry has grown rapidly in recent years,resulting in a rising number of ships on the sea.The tradtional ship’s route planning mainly relies on the experienced crew to draw manually on paper charts,which is inefficient and tedious and requires relevant personal to equip with high skills,therefore,it is necessary to enhance the intelligence of ship’s route planning to reduce the burden of the crew.In this work,we proposed a method to train an improved long and short-term memory artificial neural network(LSTM)by mining historical ship trajectories from AIS data as a dataset,so that it can automatically generate routes in different sea areas and recommend routes from multiple factors.The content in this work focuses on the following three points:(1)For the current problem of low quality of AIS data due to the excessive amount of AIS data and the high number of duplicate,redundant,missing and erroneous data,this work analyses the AIS data and improves the quality of AIS data by means of cleaning the data and repairing or interpolating the missing data.For the areas where route data is scarce,this work thins the routes with longer lengths to facilitate the training of the model.(2)It is proposed that a route planning method combining K-means algorithm and improved LSTM algorithm.In order to obtain the historical routes under different paths,the K-means algorithm is used to cluster the target sea area.Then the sorted dataset is put into the improved model for training,and the error of the model is found to be reduced comparing with that before the improvement.(3)Generate data of ships sailing on different paths using(2)method and analyze the data;Generate routes for different types of ships,and analyze and plan routes based on sailing time,distance,fuel consumption comparison,navigation safety,and the possibility of navigation congestion.Finally,the applicability of the model in this paper was discussed,and the results showed that the route planning method in this paper is applicable to different sea areas.