QoS-Aware Beam Load Balancing for LEO Satellites: An Attractor Selection Model With Priority Queuing

Multi-beam low Earth orbit (LEO) satellite networks face critical load imbalance due to dynamic topology and uneven traffic distribution, severely degrading transmission quality. To address this, we propose a multidimensional attractor selection load balancing (MAS-LB) algorithm that integrates attractor selection with priority queuing for quality of service (QoS)-aware dynamic beam resource allocation. First, a non-preemptive priority single-server queuing (NPPSSQ) model is established to differentiate service requirements and guarantee QoS. Second, a closed-form expression for average delay is derived via queuing theory, enabling precise latency prediction under dynamic loads. Third, the proposed MAS-LB algorithm dynamically allocates beam resources by jointly exploiting user demands, channel states, and beam conditions. Moreover, the steady-state condition of the NPPSSQ model guarantees Lyapunov- like convergence of the attractor dynamics, certifying the stability of this otherwise heuristic load-balancing design. Simulations demonstrate that the proposed MAS-LB with NPPSSQ scheme reduces average end-to-end delay by up to 32.5% compared to genetic algorithm (GA) and 23.1% compared to baseline MAS under 300% load, while improving normalized throughput by 97.5% and 79.2%, respectively. Moreover, it achieves 98.2% QoS compliance for high-priority traffic under 200% overload, approaching learning-based methods without training overhead. This work establishes a hybrid optimization paradigm bridging bio-inspired attractor selection and priority queuing for dynamic resource allocation in 6G-integrated space-air-ground networks.