Efficient concurrent GHZ state distribution in quantum networks

Distributing multipartite entanglement, particularly Greenberger-Horne-Zeilinger (GHZ) states, across quantum networks presents unique challenges due to the stochastic nature of link-level entanglement generation and the resource constraints of quantum memories. Existing approaches typically rely on static pre-planning and single-request processing, which fail to adapt to real-time network conditions and handle concurrent demands efficiently. This paper introduces a novel two-level framework that addresses these limitations: (1) a Multi-Armed Bandit (MAB) approach using Thompson Sampling to dynamically allocate entanglement-generation capacity across network links based on observed success rates, and (2) a concurrency-aware scheduling system that manages multiple simultaneous GHZ requests under memory constraints. Our evaluation demonstrates that this integrated approach achieves a considerable reduction in GHZ state generation latency compared to static allocation strategies. In non-stationary network environments, our MAB-based algorithm automatically shifts resource allocation, maintaining optimal throughput where static approaches fail. For concurrent requests, our scheduling achieves considerably faster performance than traditional FIFO approaches by minimizing resource conflicts between concurrent requests. The framework’s ability to balance exploration and exploitation through Thompson Sampling, combined with its resource-guided scheduling, provides a robust foundation for scalable quantum networking applications across dynamically changing environments.