LLM-guided DRL for Multi-tier LEO Satellite Networks with Hybrid FSO/RF Links

TL;DR

We address reliable downlink in a three-tier LEO-HAP-ground network with hybrid FSO/RF links under high mobility, formulating a joint optimization of end-to-end rate and handover frequency. The authors propose LTQC-DAM, a DRL framework that combines dynamic action masking to prune invalid satellite actions and an LLM-guided controller to adapt hyperparameters during training, built on a truncated quantile critic backbone. The method explicitly handles cross-tier dependencies by enforcing a flow constraint $\sum_{i} R_{RF_i}(t) \le R_{FSO}(t)$ and optimizing subcarrier allocations, satellite selection, and user choices within an MDP. Simulations show faster convergence and improvements in $f_1$ (throughput) and $f_2$ (handover minimization) over baselines, with DeepSeek providing the best LLM-driven tuning.

Abstract

Despite significant advancements in terrestrial networks, inherent limitations persist in providing reliable coverage to remote areas and maintaining resilience during natural disasters. Multi-tier networks with low Earth orbit (LEO) satellites and high-altitude platforms (HAPs) offer promising solutions, but face challenges from high mobility and dynamic channel conditions that cause unstable connections and frequent handovers. In this paper, we design a three-tier network architecture that integrates LEO satellites, HAPs, and ground terminals with hybrid free-space optical (FSO) and radio frequency (RF) links to maximize coverage while maintaining connectivity reliability. This hybrid approach leverages the high bandwidth of FSO for satellite-to-HAP links and the weather resilience of RF for HAP-to-ground links. We formulate a joint optimization problem to simultaneously balance downlink transmission rate and handover frequency by optimizing network configuration and satellite handover decisions. The problem is highly dynamic and non-convex with time-coupled constraints. To address these challenges, we propose a novel large language model (LLM)-guided truncated quantile critics algorithm with dynamic action masking (LTQC-DAM) that utilizes dynamic action masking to eliminate unnecessary exploration and employs LLMs to adaptively tune hyperparameters. Simulation results demonstrate that the proposed LTQC-DAM algorithm outperforms baseline algorithms in terms of convergence, downlink transmission rate, and handover frequency. We also reveal that compared to other state-of-the-art LLMs, DeepSeek delivers the best performance through gradual, contextually-aware parameter adjustments.

Figures (5)

• Figure 1: The considered multi-tier hybrid satellite downlink communication system composed of a LEO satellite constellation, an HAP, and multiple ground user clusters.
• Figure 2: Framework of the LTQC-DAM algorithm for the considered multi-tier hybrid satellite downlink communication system.
• Figure 3: Convergence comparison of different algorithms.
• Figure 4: Average optimization objective performance comparison of different algorithms, including downlink transmission rate ($f_1$) and satellite handover frequency ($f_2$).
• Figure 5: Visualization of hyperparameter adaptation patterns across different LLMs during training.