Enhancing Sustainability in HAPS-Assisted 6G Networks: Load Estimation Aware Cell Switching

TL;DR

This paper tackles the challenge of traffic load estimation in HAPS-assisted 6G networks, proposing spatial and temporal prediction methods to improve energy-efficient cell switching in heterogeneous networks.

Contribution

It introduces novel spatial and temporal load estimation techniques, including multi-level clustering and LSTM, to enhance accuracy in sleep mode base stations.

Findings

MLC and LSTM methods outperform other approaches in load estimation accuracy.

Spatial and temporal estimation significantly improve energy-saving strategies.

Real-world data validates the effectiveness of proposed methods.

Abstract

This study introduces and addresses the critical challenge of traffic load estimation in cell switching within vertical heterogeneous networks. The effectiveness of cell switching is significantly limited by the lack of accurate traffic load data for small base stations (SBSs) in sleep mode, making many load-dependent energy-saving approaches impractical, as they assume perfect knowledge of traffic loads, an assumption that is unrealistic when SBSs are inactive. In other words, when SBSs are in sleep mode, their traffic loads cannot be directly known and can only be estimated, inevitably with corresponding errors. Rather than proposing a new switching algorithm, we focus on eliminating this foundational barrier by exploring effective prediction techniques. A novel vertical heterogeneous network model is considered, integrating a high-altitude platform station (HAPS) as a super macro base station (SMBS). We investigate both spatial and temporal load estimation approaches, including three spatial interpolation schemes, random neighboring selection, distance based selection, and multi level clustering (MLC), alongside a temporal deep learning method based on long short-term memory (LSTM) networks. Using a real world dataset for empirical validation, our results show that both spatial and temporal methods significantly improve estimation accuracy, with the MLC and LSTM approaches demonstrating particularly strong performance.