Level 4 autonomous ship situational awareness system architecture and sensor trust management protocol

The relevance of this study is determined by the fact that, for Level 4 autonomous ships, traditional situational awareness architectures do not comply with Rule 5 of the International Regulations for Preventing Collisions at Sea concerning continuous lookout by all available means, since such ships do not distinguish between random noise, sensor degradation, and cyberattacks. This paper proposes a Level 1 situational awareness architecture consistent with the Endsley model. Sensors are divided into a continuously active core (radar, satellite navigation systems, Automatic Identification System, optical and thermal imaging cameras, microphones, and other systems) and specialized modules (LiDAR, Doppler speed log, sonar). A Sensor Trust Manager with a quality metric and a trust management protocol for navigational information sensors have been developed, combining noise level, deviation from the predicted value, cyberattack (spoofing) detection, and packet loss. The Trust Manager operates a six-state automaton (NOMINAL, DEGRADED, SUSPECT, ISOLATED, RECOVERING, FAILED) with transition hysteresis, ensuring resistance to short-term interference without false sensor isolation. Trust Managers ensure the isolation of unreliable sources before data enter the fusion filters. Thus, only physically reliable and time-consistent data are supplied to complex fusion algorithms (Kalman filters, trackers), which is critical for preventing “contamination” of the target vessel state assessment. The proposed architecture combines compliance with Rule 5, scenario adaptation through the division of sensors into basic and specialized groups, and cyber resilience (spoofing) through signal evaluation based on quality criteria, creating a theoretical basis for certification of Level 4 autonomous marine surface ships in accordance with the requirements of the Code of Safety for Autonomous Surface Ships .