A Decentralized Epidemic LoRa Mesh V2X Architecture for Fleet-Wide Situational Awareness and Opportunistic Coast Guard Alerting in Artisanal Maritime CPS

Abstract

Artisanal fishing fleets in the Bay of Bengal operate with little or no continuous cellular or satellite coverage, and Automatic Identification System (AIS) or VSAT-class equipment is not economically viable for 10-to-15-meter wooden vessels. We propose a decentralized, edge-intelligent maritime cyber-physical system (CPS) in which every vessel senses and locally classifies its own position, kinematic (oceanic) distress state, mechanical (in-boat) anomaly state, and nearby-intruder status using unsupervised edge machine learning. Rather than routing telemetry to a fixed gateway, vessels exchange and replicate a shared fleet-wide telemetry table over a decentralized vessel-to-vessel (V2X) LoRa mesh using an epidemic (gossip) protocol, so that every vessel eventually holds a synchronized copy of the fleet's last-known state. Because vessels routinely transit toward and away from shore, any vessel that comes within cellular or fixed-gateway range opportunistically uploads its cached fleet-wide table to a coast guard backend, giving near-continuous composite situational awareness without requiring any single vessel to hold a persistent connection. We validate the onboard unsupervised anomaly detectors and the near-field radar physics governing intruder detection on low radar-cross-section wooden hulls through synthetic, physics-grounded simulation, and evaluate the proposed epidemic mesh against a fixed-gateway multi-hop baseline under one shared, corrected two-ray/shadow-fading link-budget model over a realistic 5--60~km offshore fishing area. Over a 6-hour mobility window, the proposed mesh achieves a 0.989--0.999 eventual delivery ratio across 20--100 vessel densities, markedly higher than the fixed-gateway baseline's instantaneous multi-hop packet delivery ratio (PDR) of 0.451--0.999 over the same range, while delivering a median distress-to-shore alert in 33.8--48.5~s.