Quantum Deep Reinforcement Learning for URLLC Satellite-Air-Ground Integrated Networks With Digital Twin Applications

In this article, we explore a maritime 6G-enhanced satellite-air-ground integrated network (SAGIN) that incorporates a unmanned aerial vehicle (UAV)-carried reconfigurable intelligent surface (UCR) relay, and low-Earth orbit (LEO) satellites equipped with mobile edge computing (MEC) facilities. The system captures dynamic maritime conditions, including ultrareliable and low-latency communication (URLLC) user mobility and UCR movements across harbor environments. The primary objective is to minimize the total system cost by jointly optimizing task offloading decisions, bandwidth allocation, local computational resource distribution, transmission power control, and caching management, while satisfying strict latency and resource constraints. To address this, we formulate a mixed-integer nonlinear programming (MINLP) problem that captures the complexity of resource optimization in the maritime 6G-enhanced SAGIN. Two quantum-enhanced deep reinforcement learning (DRL) algorithms, namely quantum-enhanced deep deterministic policy gradient (QEDDPG) and quantum-enhanced proximal policy optimization (QEPPO), are proposed to solve the formulated MINLP problem. Moreover, higher order quantum feature encoding and quantum neural networks (QNNs) are utilized to accelerate learning and enhance decision-making. Simulation results demonstrate that QEDDPG and QEPPO significantly outperform conventional DRL methods by achieving lower system costs and more efficient resource allocation. These findings show the potential of quantum-driven reinforcement learning for enabling scalable, efficient, and intelligent resource management in future 6G-enhanced SAGINs.