Prediction-Driven Assessment of Multi-Ship Traffic Pressure and Maritime Traffic Situation

In increasingly complex navigation environments, maritime traffic supervision needs to look beyond the instantaneous collision risk of individual-ship pairs. A multi-ship scene may become difficult to monitor because of vessel aggregation, spatial compression, encounter urgency, and inconsistent motion states. To support proactive Vessel Traffic Services (VTS), this study proposes a prediction-driven framework for assessing multi-ship traffic pressure by combining AIS-based short-term motion prediction with a Spatio-Temporal Encounter Traffic Pressure Index (ST-TPI). In the proposed framework, cleaned and resampled AIS trajectories are used to train an LSTM model for short-term vessel motion prediction. The predicted vessel states are then synchronized into future multi-ship traffic snapshots over a 30 min horizon, and ST-TPI is used to evaluate traffic pressure at the ship-pair, individual-ship, regional, and scene levels. Different from conventional collision-risk or traffic-complexity methods, the proposed framework focuses on how future traffic pressure forms, changes, and is transferred among vessels and vessel pairs. The method was tested using five typical multi-ship scenarios and a real-waterway case in the western precautionary area of the Laotieshan Channel. The prediction results showed stable short-term forecasting performance with low meter-level position errors under the observation-updated rolling evaluation, providing a basis for future multi-ship snapshot generation. The typical scenarios revealed different pressure-evolution patterns, including low-pressure persistence, temporary compression and release, delayed crossing pressure, complex interaction release, and High-level pressure formation. The real-waterway case further showed low and Low-medium pressure fluctuations, local pressure peaks, pressure release, and pressure-source transfer under practical AIS conditions. Prediction-error perturbation analysis indicated that the main high-pressure vessel pairs and pressure-level interpretations remained stable under tested position perturbations. Consistency analysis further showed that ST-TPI scene pressure was significantly correlated with conventional CRI-based encounter-risk indicators. These results indicate that the proposed framework can provide interpretable information on future pressure-evolution and dominant pressure sources, supporting proactive monitoring, early warning, and traffic organization in complex waterways, and contributing to a safer maritime traffic environment.