Contemporary maritime domains such as offshore energy, infrastructure inspection, and environmental assessment are characterised by increasingly data-intensive operations. While sensor technologies continue to evolve in resolution and throughput, downstream processing remains a limiting factor, with analytical workflows often requiring multiples of the original acquisition time. This latency inhibits scalability and delays critical decision-making.
This study presents a modular edge computing architecture designed to support real-time, multi-sensor data processing in subsea environments. The framework was initially validated through long-term studies (2018–2024) on subsea pipeline inspection campaigns, employing fused optical and geometric data streams to perform in situ anomaly detection and reporting. These foundational results demonstrate the feasibility of deploying complex analytics proximate to the sensor source.
The investigation extends to the integration of non-visual sensing modalities, with a focus on synthetic aperture sonar (SAS). The architecture supports the decomposition of processing pipelines into discrete stages, enabling sensor-specific workflows to be implemented at the edge. This not only reduces the dependency on high-bandwidth data transmission but also facilitates adaptive behaviour and autonomy in platform navigation and task execution.
Empirical results from SAS-based surveys are used to evaluate system performance across dimensions including latency, accuracy, and processing load. The broader implications of this approach are discussed in relation to enabling scalable, mission-driven autonomy and transitioning from post hoc analysis models to real-time, embedded decision-support systems within the maritime domain.
