Fast Trajectory Forecasting With Automatic Identification System Broadcasts

• Published in: 2022 Sensor Signal Processing for Defence Conference (SSPD) pp. 1 - 5
• Main Authors: Wang, Yicheng, Uney, Murat
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2022
• Subjects: Computational modeling; Prediction algorithms; Predictive models; Signal processing; Signal processing algorithms; Stochastic processes; Trajectory
• DOI: 10.1109/SSPD54131.2022.9896218

This work proposes a fast trajectory forecasting algorithm to use with automatic identification system (AIS) broadcasts of vessels. The algorithm involves fast sub-optimal model parameter estimation from AIS messages and the computation of Gaussian location predictions for a series of future timestamps. The underlying trajectory model is a stochastic process that uses six parameters to generate near-constant velocity trajectories. These parameters include the desired cruise heading and speed of the vessel and velocity standard deviations along the heading direction and its perpendicular complement. We demonstrate the performance of our approach using a real AIS data set.