AISAI

Automatic Identification System (AIS) data has transformed maritime operations over the past two decades by enabling real-time situational awareness, vessel tracking, and collision avoidance. Mandated by the International Maritime Organization (IMO) since 2000, AIS systems automatically broadcast vital navigational and vessel-related information, such as position, speed, course, heading, vessel type, and destination, to nearby ships and coastal authorities. This dissertation presents a unified framework for the intelligent collection, processing, storage, and analysis of AIS data, addressing both the scalability and reliability challenges associated with real-time maritime data streams.

Building upon real-world deployments in the Eastern Mediterranean, the research contributes a suite of integrated algorithms and services designed to enhance maritime informatics. The proposed framework consists of three core components: AIS data cleaning, vessel Estimated Time of Arrival (ETA) prediction, and vessel trajectory forecasting. First, the system applies advanced data preprocessing techniques to remove redundancies, resolve inconsistencies, and impute missing values in AIS messages. Second, machine learning models are trained to provide accurate, dynamic ETA predictions, allowing ports and shipping stakeholders to optimize resource allocation and berth scheduling and reduce vessel idle times. Third, deep learning-based sequence models are used to forecast vessel trajectories over short and long horizons, supporting proactive traffic management, navigational safety, and operational planning.

By integrating these capabilities, this work advances the development of intelligent, scalable, and environmentally aligned maritime decision-support systems. The findings have practical implications for port authorities, shipping companies, and regulatory bodies, promoting safer, more efficient, and sustainable maritime logistics.