Towards Stability and Utility Maximization in Adversarial Quantum Networks under Bandit Feedback (Invited)

This paper addresses the dual challenges of stabilizing entanglement scheduling and maximizing communication fidelity in multi-hop quantum networks under adversarial dynamics. Unlike prior works relying on stationary assumptions or perfect pre-decision knowledge, we consider scenarios where entanglement generation rates and task arrival patterns vary arbitrarily over time, with only bandit feedback available postscheduling. We first propose Q–NSO, achieving stability for quantum networks under arbitrary time-varying entanglement request arrival rates and bandit feedback. Equipped with the key components of Q-NSO, we propose Q-UMO, a learning-based framework towards maximal utility by determining the entanglement request arrivals and quantum link allocation distribution. Integrating adversarial bandit convex optimization, an online learning algorithm, and Lyapunov drift-plus-penalty analysis, theoretical analysis demonstrates that Q-UMO ensures network stability with bounded queue lengths while achieving a sub-linear regret bound against the reference policy. This work provides a robust foundation for scalable quantum internet applications, balancing stability and performance in adversarial environments with limited feedback.