In this paper, we model the trajectory of sea vessels and provide a servicethat predicts in near-real time the position of any given vessel in 4', 10',20' and 40' time intervals. We explore the necessary tradeoffs betweenaccuracy, performance and resource utilization are explored given the largevolume and update rates of input data. We start with building models based onwell-established machine learning algorithms using static datasets andmulti-scan training approaches and identify the best candidate to be used inimplementing a single-pass predictive approach, under real-time constraints.The results are measured in terms of accuracy and performance and are comparedagainst the baseline kinematic equations. Results show that it is possible toefficiently model the trajectory of multiple vessels using a single model,which is trained and evaluated using an adequately large, static dataset, thusachieving a significant gain in terms of resource usage while not compromisingaccuracy.
展开▼