First Advisor

John Lipor

Term of Graduation

Fall 2020

Date of Publication

12-8-2020

Document Type

Thesis

Degree Name

Master of Science (M.S.) in Electrical and Computer Engineering

Department

Electrical and Computer Engineering

Language

English

Subjects

Ships -- Automatic identification systems, Underwater acoustics, Neural networks (Computer science), Acoustic localization

DOI

10.15760/etd.7500

Physical Description

1 online resource (x, 56 pages)

Abstract

Maritime vessel position coordinates are important information for maritime situational planning and organization. A better estimate of future locations of the maritime vessels, from their current locations, can help maritime authorities to make planned decisions, which can be helpful to avoid traffic congestion and longer waiting times. This thesis develops a method for estimating future locations of the vessels using their current and previous locations and other data.

The motivating scenario for this work is that of determining the future locations of the vessels based on their current location and previous locations for accurate modelling of underwater acoustic noise. As ambient noise levels can be reliably calculated based on ship traffic, the location of ships have a significant impact on noise level estimates. Methods to predict future locations of the vessels can be classified into two types, one which tries to model the physical behaviour of the ship, and methods which are data driven. The problem with the first approach is that there is a large number of parameters such as ocean current, weather, wind speed, vessel speed, and direction to develop an accurate physical model of the ship. Incorporating all the parameters in the physical model becomes intractable. The advantage of a data driven approach is that all the parameters, which may decide the physical behaviour of the vessel, are incorporated through the data.

We have developed two methods leveraging historical Automatic Identification System (AIS) data to come up with a model to predict the future location of the vessels. The first method uses a simple mathematical construct called the Markov model for the prediction task, which computes the transition probability matrix for the given Region of Interest (ROI), then the computed transition probability matrix is used to make predictions. The second method uses a neural network-based technique called Long-Short Term Memory (LSTM), which incorporates side information along with location data to model the behaviour of the vessel.

We evaluated the performance of both the methods in multiple regions near the US coasts. We show empirically that the Markov model-based method, for a smaller region (22,500 km2), is able to predict the location of the vessel with an average error of 11.7 km at an interval of 4 hours. We show empirically that the LSTM-based method, which is more general and applicable for the bigger region as well, is able to predict the location of the vessels with an average error of 12.18 km at an interval of 4 hours. The ability of LSTM-based method to incorporate other information makes it well suited for the task of vessel trajectory prediction.

Considering future work, we propose a combination of two methods, as well as other data sources like weather and complex neural network models that may be worth further investigation. This thesis's contributions are detailed investigation and analysis of the Markov model approach and LSTM-based approach on different regions near the US coastal area. We provide a detailed analysis of the effect of adding different features to LSTM-based model. Additionally, we contribute a method for trajectory prediction for multiple timestamps in the future, which is simple and has not been tried in the current state of the art to the best of our knowledge.

Rights

In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

Persistent Identifier

https://archives.pdx.edu/ds/psu/34592

Available for download on Wednesday, December 08, 2021

Share

COinS