Two-Phase Cell Switching in 6G vHetNets: Sleeping-Cell Load Estimation and Renewable-Aware Switching Toward NES
A new AI-powered cell switching strategy for 6G networks uses LSTM predictors to achieve 1% error and 23% energy savings.
A team of researchers has introduced a novel two-phase AI framework designed to make future 6G networks dramatically more energy-efficient. The core challenge in 6G's complex vertical heterogeneous networks (vHetNets)—which mix macro, small, and aerial base stations—is intelligently turning off (or "sleeping") underused cells without disrupting service. The first phase tackles the critical prerequisite of estimating the potential traffic load on these sleeping small cells. The researchers propose three AI/data-driven methods with varying data needs: a simple distance-based estimator, a multi-level clustering approach for limited data, and a sophisticated Long Short-Term Memory (LSTM) neural network for full historical data. On real Milan call records, the LSTM model excelled, achieving a remarkably low Mean Absolute Percentage Error (MAPE) of below 1%.
In the second phase, these accurate load estimates feed into a renewable-aware cell switching strategy. This system explicitly models solar-powered small base stations (SBSs) across three realistic operational scenarios with hybrid grid/renewable setups. By dynamically adapting switching decisions based on real-time renewable energy availability and battery storage conditions, the framework moves beyond theoretical models. The result is a robust and practical system for real-world deployment. Simulations show the most effective scenario delivers up to 23% in Network Energy Saving (NES) compared to traditional cell switching, all while maintaining quality of service. This work represents a significant step toward sustainable 6G radio access network (RAN) operations, directly addressing the massive energy consumption concerns of next-generation telecommunications.
- Phase I uses an LSTM model to predict sleeping cell traffic with under 1% error (MAPE), a prerequisite for safe cell shutdowns.
- Phase II integrates these predictions into a renewable-aware switching strategy, modeling solar-powered cells for realistic deployment.
- The full framework achieved up to 23% Network Energy Saving (NES) in simulations using real Milan call data, without compromising service quality.
Why It Matters
It provides a practical AI blueprint to drastically reduce the massive energy footprint of future 6G networks, a key operational and environmental challenge.