The Prediction Model For Ship Trajectory Based On AIS Data

With the economic globalization and the expansion of the scale of international trade,the maritime transportation industry is developing rapidly,and ships are improving in the direction of large-scale and high-speed ships.The number of ships and the density of maritime traffic,as well as the types and loading of dangerous goods continue to rise.Although the increase in the number of ships and shipping routes makes maritime trade more and more prosperous,at the same time,it will congest the waterways where ships frequently travel and increase the load.Correspondingly,the ship’s own problems and human factors will also lead to an increase in accidents,which seriously threaten the safety of the lives and property of the people on the ship.At the same time,it poses new challenges for ship collision avoidance,port traffic management and navigation safety.Therefore,predicting the trajectory of ships in advance is of great significance to ensure the safety of maritime traffic on ships.As AIS historical data contain a large amount of ship traffic behavior information,important ship traffic behavior characteristics can be obtained by analyzing AIS data.This dissertation uses a large amount of historical AIS data,analyzes the characteristics of the data and uses deep learning theory to study the algorithms suitable for ship trajectory prediction,and build a ship trajectory prediction model based on the GRU-LR hybrid model.The research content of this dissertation is as follows:1.This dissertation analyzes the AIS data and selects ships that meet the model training.And deal with vacancy values,abnormal values,and extract ship trajectory data that meets the model training.And use the random forest algorithm to rank the importance of the AIS data features,and select the data features that have a greater impact on the model target.2.This dissertation establishes the prediction model for ship trajectory based on GRU-LR.When processing time series,because the GRU model can capture the long-term dependence in the time series data,the GRU(Gated Recurrent Unit)and LR(Linear Regression)are combined to construct a GRU-LR hybrid model.3.The simulation experiment verifies that the GRU-LR trajectory prediction model is correct and effective,with better training effects and smaller errors.