Queue-Informed Neural Network Model for Estimating Queueing Delay in PON-Based Aggregation Networks

Future carrier networks will need to aggregate traffic from widely distributed sensor nodes while meeting stringent service requirements in industrial and logistics applications. As network scales expand, both deterministic latency assurance and cost-effectiveness become increasingly important. Technologies with optical transmission and aggregation capabilities, such as Passive Optical Networks (PONs) with switching functions, offer cost-effective solutions. However, multi-node traffic aggregation introduces queueing delays, which must be accurately estimated to meet performance requirements. Traditional machine learning approaches for queueing delay estimation often rely heavily on large labeled datasets, which are expensive and impractical to collect in large-scale environments. To overcome this limitation, we propose a Queue-Informed Neural Network (QINN) model that directly incorporates governing equations derived from queueing theory into the neural network training process. By embedding queueing principles into the learning model, the proposed QINN improves the accuracy of queueing delay estimation without requiring extensive datasets. The contributions of this study are threefold: first, we design a novel loss function that combines data-driven error minimization with theoretical consistency based on queueing models; second, we propose the QINN, which improves queueing delay estimation accuracy under limited training data conditions by embedding queueing-theory-based constraints into the learning objective; and third, we model and evaluate the computational cost of QINN to confirm its feasibility for deployment in real-world environments with constrained processing resources. These findings demonstrate that integrating queueing-theory-based formulations into learning models enhances estimation performance while maintaining computational efficiency even in complex multi-node aggregation scenarios such as PON combined with Ethernet switches.