Unsupervised anomaly detection for underwater gliders using generative adversarial networks

An effective anomaly detection system is critical for marine autonomous systems operating in complex and dynamic environments to reduce operational costs achieve concurrent large-scale fleet deployments. However, developing an automated fault remains challenging several reasons including limited data transmission via satellite services. Currently, most systems, such as underwater gliders, rely on intensive analysis by pilots. This study proposes unsupervised using bidirectional generative adversarial networks guided assistive hints with time series collected multiple sensors. In this study, the a of gliders trained two healthy deployment datasets tested other nine selection vehicles range locations environmental conditions. The successfully applied detect anomalies test deployments, which include different types well behaviour. Also, sensitivity decimation settings suggests proposed robust Near Real-Time gliders.