Navigational risk assessment of inland waters based on bi-directional PSO-LSTM algorithm and ship maneuvering characteristics

Inland waterways are characterized by various bends, complex intersections and high traffic density. This complex inland navigation environment brings challenges to vessel traffic service (VTS) in managing inland water traffic. We propose a method to assess navigational risk of inland waters based on bi-directional PSO-LSTM algorithm and ship maneuvering characteristics. First, a bi-directional PSO-LSTM based ship trajectory restoration model is established to repair the Automatic identification System (AIS) data. Subsequently, ship behavioral characteristics are extracted from these repaired AIS data, based on which the model of ship navigation risk assessment in inland waters is established using window approach. Several case studies are conducted to verify this proposed method. The promising results attest that this proposed bi-directional PSO-LSTM algorithm outperforms the traditional neural network and trajectory restoration models. In addition, this navigational risk assessment of inland waters can effectively identify complex waters with high navigational risks, which are consistent with the actual situation. This research helps support VTS in regulating ship traffic in inland waters, thereby ensuring the safety of ship navigation on inland waterways.