In the ever-evolving world of maritime technology, a novel solution has emerged to tackle a persistent challenge: the unreliable data stream from the Automatic Identification System (AIS). Hee-jong Choi, from the Department of Maritime Transportation System at Mokpo National Maritime University in South Korea, has developed a framework called BAARTR, which stands for Boundary-Aware Adaptive Regression for Kinematically Consistent Vessel Trajectory Reconstruction. This innovative approach aims to improve the accuracy of vessel trajectory reconstruction, a critical task for navigation, monitoring, and safety at sea.
AIS, a system used for tracking and identifying vessels, often suffers from data loss and irregular reporting intervals. This inconsistency can compromise the reliability of downstream tasks, including trajectory reconstruction. Choi’s BAARTR framework addresses this issue by operating solely on time, latitude, and longitude inputs. It explicitly enforces boundary velocities derived from raw AIS data, adaptively selecting a velocity-estimation strategy based on the AIS reporting gap. For short intervals, it uses central differencing, while for longer or irregular gaps, it employs a hierarchical cubic velocity regression with a quadratic acceleration constraint to iteratively refine endpoint slopes.
These boundary slopes are then incorporated into a clamped quartic interpolation at a 1-second resolution, effectively suppressing overshoots and ensuring velocity continuity across segments. Choi explains, “BAARTR achieves superior reconstruction accuracy while maintaining strictly linear time complexity. It consistently achieved the lowest median Root Mean Square Error (RMSE) and the narrowest Interquartile Ranges (IQR), producing visibly smoother and more kinematically plausible paths—especially in high-curvature turns where standard geometric interpolations tend to oscillate.”
The commercial impacts of this research are significant. Accurate trajectory reconstruction is crucial for various maritime sectors, including Maritime Autonomous Surface Ships (MASS) and Vessel Traffic Services (VTS). It also plays a vital role in accident reconstruction and multi-sensor fusion. By reducing reliance on large training datasets, BAARTR offers a lightweight, extensible foundation for these applications.
Choi’s sensitivity analysis shows that BAARTR performs stably with a modest training window (n ≈ 16) and minimal regression iterations (m = 2–3). This efficiency makes it a practical solution for real-world maritime environments. As Choi notes, “By reducing reliance on large training datasets, BAARTR offers a lightweight, extensible foundation for post-processing in Maritime Autonomous Surface Ship (MASS) and Vessel Traffic Service (VTS), as well as for accident reconstruction and multi-sensor fusion.”
The research was published in the Journal of Marine Science and Engineering, a testament to its scientific rigor and potential impact. For maritime professionals, BAARTR represents a significant advancement in trajectory reconstruction technology, promising improved safety, efficiency, and reliability in maritime operations. As the maritime industry continues to embrace digital transformation, innovations like BAARTR will play a pivotal role in shaping the future of sea travel and trade.