Deep Learning Applications in Vessel Dead Reckoning to Deal with Missing Automatic Identification System Data

Abstract

This research introduces an online system for monitoring maritime traffic, aimed at tracking vessels in water routes and predicting their subsequent locations in real time. The proposed framework utilizes an Extract, Transform, and Load (ETL) pipeline to dynamically process AIS data by cleaning, compressing, and enhancing it with additional attributes such as online traffic volume, origin/destination, vessel trips, trip direction, and vessel routing. This processed data, enriched with valuable details, serves as an alternative to raw AIS data stored in a centralized database. For user interactions, a user interface is designed to query the database and provide real-time information on a map-based interface. To deal with false or missing AIS records, two methods, dead reckoning and machine learning techniques, are employed to anticipate the trajectory of the vessel in the next time steps. To evaluate each method, several metrics are used, including R squared, mean absolute error, mean offset, and mean offset from the centerline. The functionality of the proposed system is showcased through a case study conducted in the Gulf Intracoastal Waterway (GIWW). Three years of AIS data are collected and processed as a simulated API to transmit AIS records every five minutes. According to our results, the Seq2Seq model exhibits strong performance (0.99 R squared and an average offset of ~1400 ft). However, the second scenario, dead reckoning, proves comparable to the Seq2Seq model as it involves recalculating vessel headings by comparing each data point with the previous one.

An Adaptive Multimodal Data Vessel Trajectory Prediction Model Based on a Satellite Automatic Identification System and Environmental Data

“…Zhang et al [29] designed a position prediction approach to increase ship position forecast accuracy in ship traffic engineering by employing the k-nearest neighbors (KNN) algorithm. Sedagha et al [22] presented a system framework for online maritime traffic monitoring aimed at the real-time tracking of vessels on waterways and the prediction of their subsequent positions. Through employing a decomposition reconstruction process and adaptive segmented error correction, Wei et al [30] developed a multi-objective heterogeneous integration approach for predicting ship motion.…”

“…However, the application scenarios of traditional methods depend on boundary conditions. Machine learning methods can improve prediction accuracy by creating complex mathematical models to simulate ship movements [22,23]. However, machine learning approaches need to collect a considerable amount of labeled data and the establishment of appropriate rules.…”

Automatic Identification System-Based Prediction of Tanker and Cargo Estimated Time of Arrival in Narrow Waterways

“…This paper does not discuss the algorithm but rather uses its output to predict ETA for vessels. For an in-depth exploration of the methodology, we recommend reviewing our previous works [50,51].…”

“…AIS data encompass details such as a vessel's location, speed, date, and time. In this section, the primary features are initially extracted from the raw data utilizing the Sedaghat [50,51] algorithm. Their approach enables the retrieval of information related to a vessel's route, velocity, direction, and trip number.…”

“…Within this module, feature extraction, identification of missing values, feature engineering, and the division of data into training and testing sets were carried out. Following the approach suggested by Sedaghat et al [50,51], the initial features were extracted from the raw data in the first step. Table 3 displays the list of features extracted from the raw data.…”