Cooperative underwater anomaly detection in Autonomous Underwater Vehicle (AUV) - Autonomous Surface Vehicle (ASV) teams

Autonomous marine systems, particularly Autonomous Underwater Vehicles (AUVs) and Autonomous Surface Vehicles (ASVs), play a crucial role in wide-area underwater monitoring tasks. This thesis examines the deployment of object detection algorithms within these systems in order to advance cooperative communications by exploiting communication-computation trade-off. The research focuses on the application of Gabor filters and Sobel filters for edge detection and texture analysis, complemented by contour detection and k-means clustering techniques. An ensemble model integrating these methods is developed to improve detection accuracy and reduce false positives, which is essential given the high operational costs of ASVs. By evaluating the performance of these algorithms in AUVs, ASVs, and distributed configurations, the study aims to determine the optimal setup for effective anomaly detection and follows the split computing paradigm in order to balance communication-computation trade-off within cooperative communication strategies. The results highlight the significant role of simplified versions of advanced algorithms in enhancing the efficiency and reliability of autonomous marine operations and analysis on different communication - computation scenarios by splitting the complexity between two communicating vehicles which contributes to the advancement of marine surveillance technologies.