Bayesian Localisation in the Underwater Environment (BLUE)

Project BLUE is a collaborative research project between the University of Liverpool and the National Oceanography Centre (NOC). The project benefits from Liverpool’s expertise in efficient Bayesian methods, distributed computing, and practical sensor processing, and NOC’s capabilities in dynamic ocean modelling and sonar sensing.

Challenges

Information processing – sensor data fusion, performance metrics, automated tracking and the fundamental properties of signal processing techniques applied to problems associated with underwater targets and acoustic signals of interest.

Optimised and adaptive sonar – technical challenges concerned with adaption for the underwater acoustic environment.

Detection, classification and localisation – Detecting objects from underwater acoustic signals, classification to establish the identity of the objects, and localisation to find the relative bearing, the range and, if possible, the depth. These processes need to work when signal-to-noise ratio is low, the environment is very cluttered, and signals of interest are very short-lived.

Aims
The project focus is on extending the state-of-the-art and so developing novel solutions to the signal processing challenges that are encountered in the underwater sensing.

Research

The approach will reflect the components of the processing chain:

Array processing: processing raw data from a single sensor array with the primary purpose of inferring the direction of arrival and (in active contexts) the range of objects;
Tracking: processing the detected angles of arrival over time scales that enable dynamic contacts to be localised and information about the objects to be refined over time;
Imaging: processing the raw data to infer information about the environment;
Feedback: adapting the parameters of the system in response to the information extracted from the other processes (both over short time scales in the guise of “control” and over long time scales in the guise of “assessment” of performance);
Simulation: ensuring that we can test the algorithms we develop in a challenging setting where ground-truth is known.

This approach will maximise the potential for the development of solutions to neighbouring problems as well as ensuring that we can both test new algorithms against meaningful baselines and identify how to exploit new algorithms that outperform existing signal processing solutions.