Answers are grounded in this paper and cite their sources.
Traditional state estimators for autonomous surface vehicles (ASVs) often rely on precise models, which can result in poor estimation accuracy in dynamic ocean environments, particularly under cyberattacks. To address this limitation, a hybrid deep learning-based estimator using only input/output data is designed that integrates a data-driven ASV model into an unscented Kalman filter (UKF) for noise suppression, employs a temporal convolutional network (TCN) with a multihead multiplicative temporal attention mechanism (MTAM) for adaptive feature extraction, and uses bidirectional gated recurrent units (Bi-GRU) to capture bidirectional temporal dependencies. This integrated approach, termed UTMB, establishes a generalization error bound through an analysis of Rademacher complexity, which provides a theoretical basis for assessing prediction reliability in resilient control. Building on the UTMB-based estimator, a resilient data-driven control method is designed for ASVs under hybrid attacks, considering both denial-of-service (DoS) attacks and false data injection (FDI) attacks. Furthermore, theoretical analysis demonstrates that the proposed intelligent control strategy ensures the stability of the tracking-error system for ASVs. Simulation results validate the effectiveness of the proposed control scheme.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: