COLLABORATION

Bath University’s IMI develops AI to map marine environments

SOURCE: Shutterstock
Bath University's Institute of Mathematical Innovation has successfully developed an AI algorithm that can automatically classify underwater environments directly from sonar measurements.

Institute for Mathematical Innovation (IMI) at the University of Bath has collaborated with an engineering company, Systems Engineering & Assessment Ltd (SEA), to successfully develop an Artificial Intelligence (AI) technique to map marine environments.

This project was initiated by the Defence and Security Accelerator (DASA), as part of the Ministry of Defense leveraging the institute’s expertise to provide mathematics, statistics and machine learning consultancy, as well as support research to develop an AI algorithm capable of automatically classifying underwater environments directly from sonar measurements.

YOU MIGHT LIKE

AI mapping of marine environments

Due to significant variation in underwater environments as a result of water temperature, salinity, and depth differences, sonar measurements are affected, thus impacting the accuracy of these readings.

Other factors that affect sonar measurements are the inherent features of these environments, which include seabed slope and composition. Seabeds can be either composed of soft clay or hard-packed sand, which will all impact sonar measurements differently.

This achievement by IMI and SEA was accomplished by first analysing the many characteristics of underwater environments and then classifying them into different types.

The research team then reviewed various techniques of applying AI to determine the most useful classification approach.

The selected method, Probabilistic Generative Modelling, was then adopted as a framework to develop three different AI algorithms for identifying underwater environments. The Probabilistic Principal Component Analysis (PPCA) model proved to be the basis for the most effective algorithm.

The research team then developed the model further through rigorous experimentation to achieve the highest possible classification accuracy. Upon acquiring the AI algorithm, the researchers tested its performance on a wide range of simulated acoustic data representative of a broad spectrum of underwater environments.

YOU MIGHT LIKE

Tests demonstrated that the PPCA algorithm could classify underwater environments from simulated sonar measurements with an average accuracy of 93%.

This result shows promise for practical application of the technique as classification accuracy remained high even when the team used a short spatial interval of the test data.

An alternative Latent Variable Gaussian Process (LVGP) model also showed strong performance with its ability to classify underwater environments with an accuracy of 96%.

Citing this achievement, technical Lead of Environmental Data Science, SEA, Marcus Donnelly said, “This project exceeded all our expectations for AI algorithms applied to the complexity of sonar in the underwater environment. We look forward to continuing our collaboration with the IMI following positive feedback from the Ministry of Defense.”

 

AI for mapping marine environments
The developed LVGP model can classify underwater environments with an accuracy of 96%.

IMI conducts interdisciplinary research through various partnerships

IMI harnesses the power of mathematics as a tool that delivers insight and solves complex real-world challenges. Through various research partnerships, the institute conducts applied mathematics research and helps to solve complex challenges in academic and private-sector research projects.

The IMI’s industry-agnostic accomplishments span the mapping of brain networks to developing machine learning controls for wave energy converters.

IMI’s research projects have far-reaching capabilities uncovering the various ways policymakers can stimulate socio-economic development and aims to deliver policy intervention evaluation tools.