Integrity-Aware Marine Navigation under Multi-Sensor Attacks Using Adaptive Descriptor Error-State Filtering
Year: 2026
Type: conference-paper
Abstract: This paper develops a marine positioning, navigation, and timing (PNT) estimator for coordinated sensor-integrity attacks targeting inertial measurement unit (IMU), magnetometer, and GNSS position measurements. The estimator places a descriptor unknown-input update inside a loosely coupled error-state navigation filter, so IMU-induced unknown-input components are suppressed at the correction step without augmenting the navigation error state. A separate biasestimation layer is then combined with per-axis fading-memory covariance learning to track nonstationary integrity-loss terms in the GNSS position and magnetometer channels. Simulations on a Marine Systems Simulator (MSS) surface-vessel model show that the adaptive covariance layer improves integrity-loss tracking and attitude estimation while maintaining comparable position accuracy relative to a non-adaptive two-stage baseline in the considered multi-sensor attack case.
Institutions: Concordia University, California Institute of Technology
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7201876604