← 研踪 PaperTrailarXiv:2611.14761 · v1
ComBO: Community Bayesian Optimization for Carrier Shutdown in HAPS-Hypercell Networks
首次提交 2026-11-11 05:37
版本更新 2026-11-11 05:37
雷达观测 2026-11-13 10:09
摘要
In this paper, we investigate carrier shutdown (CS) in high-altitude platform station (HAPS)-Hypercell-assisted networks, leveraging network-wide information to optimize both coverage macro cells and capacity cells CS. To address the high dimensionality of this problem, we propose ComBO, a community-aware Bayesian optimization framework that exploits user equipment (UE) reference signal received power (RSRP) reports to construct a fallback graph representing the underlying network and identify groups of strongly coupled cells. The resulting communities decompose the original optimization problem into lower-dimensional subproblems, which are sequentially optimized using trust region Bayesian optimization (TuRBO) and refined through a final joint search. Evaluation in a high-fidelity 3GPP-based system-level environment shows that ComBO consistently outperforms legacy CS strategies and overcomes the scalability limitations of global Bayesian optimization (BO). Considering normalized terrestrial power ComBO achieves nearly 84% savings while preserving quality of service. These results demonstrate the potential of community-aware decomposition as a scalable approach to large-scale network optimization.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-11 05:37 · 最新观测版本
ComBO: Community Bayesian Optimization for Carrier Shutdown in HAPS-Hypercell Networks
In this paper, we investigate carrier shutdown (CS) in high-altitude platform station (HAPS)-Hypercell-assisted networks, leveraging network-wide information to optimize both coverage macro cells and capacity cells CS. To address the high dimensionality of this problem, we propose ComBO, a community-aware Bayesian optimization framework that exploits user equipment (UE) reference signal received power (RSRP) reports to construct a fallback graph representing the underlying network and identify groups of strongly coupled cells. The resulting communities decompose the original optimization problem into lower-dimensional subproblems, which are sequentially optimized using trust region Bayesian optimization (TuRBO) and refined through a final joint search. Evaluation in a high-fidelity 3GPP-based system-level environment shows that ComBO consistently outperforms legacy CS strategies and overcomes the scalability limitations of global Bayesian optimization (BO). Considering normalized terrestrial power ComBO achieves nearly 84% savings while preserving quality of service. These results demonstrate the potential of community-aware decomposition as a scalable approach to large-scale network optimization.
查看此版本解读 →查看此版本原文 ↗