Multi-Forecast Model Predictive Control for Robust Topic Management in Maritime Publish-Subscribe Networks

Future maritime networks will support coordination and command-and-control activities for a diverse set of crewed and uncrewed systems. Publish/subscribe message-delivery offers an alternative to point-to-point messaging that decouples information publishers and subscribers. Publish/subscribe messaging offers increased robustness to intermittent connectivity and naturally supports node mobility. This work considers a publish/subscribe architecture that uses a specialized set of nodes, termed brokers, to manage the delivery of information from publishers to subscribers. We develop a robust policy for defining the topic-to-broker assignment, thereby defining topic management responsibilities. Our policy seeks a fair assignment of topics to brokers that discourages assignments that overload individual brokers. Additionally, our policy encourages topic assignments to remain unchanged whenever possible to reduce the control-messaging overhead among brokers. The performance of the proposed policy is illustrated via numerical tests on a simulated scenario.