Predicting Ship Speed in the Inland Waterways with LSTM Network

Abstract

Ship speed prediction is essential for inland waterway traffic management. However, current studies are either focused on the ocean ship’s speed which has different patterns compared to the inland waterway ships’ speed or requires physical data, i.e., engine revolution and ship Resistance, which is difficult to obtain in practical applications. This paper proposes a novel LSTM based algorithm for accurate inland waterway ship speed prediction. The environmental factors, i.e., the water speed and the water level, as well as the ships’ historical speed information, are combined by the features fusion method, and then the feature variables are put into the LSTM model. The model is easily trained by AIS data collected from the inland waterway ships. One thousand ships’ AIS data collected from Sichuan Province are adopted to test the effectiveness and efficiency of the proposed model. The experiment results confirm the proposed model can handle the inland waterway ship’s shorts-term speed prediction.