Strategy-Aware Federated Split Decision Transformer for Energy-Efficient Cellular Networks

Abstract

As Fifth Generation (5G) networks evolve toward Sixth Generation (6G), base station energy consumption poses a critical challenge for sustainable wireless infrastructure, demanding intelligent and adaptive resource management strategies. Reinforcement learning (RL) offers a promising approach for such dynamic optimization problems, yet traditional online RL approaches lack scalability due to costly and risky real-time interactions with live networks, while centralized training exposes sensitive operational data in multioperator deployments. To address these issues, we propose the Strategy-Aware Federated Split Decision Transformer (SA-FSDT) for privacy-preserving energy optimization in heterogeneous cellular networks. Our framework overcomes the spatial coordination limitations of existing methods by introducing a strategy-aware cross-attention mechanism. Specifically, small base stations compute local embeddings that are transmitted to a macro base station, which utilizes a hierarchical attention strategy to dynamically balance global network coordination with local capacity optimization. By transforming offline reinforcement learning into supervised sequence modeling, SA-FSDT eliminates risky online exploration while safeguarding proprietary information across operators. Extensive simulations demonstrate that SA-FSDT significantly reduces power consumption while simultaneously improving QoS satisfaction, and exhibits robust cross-density generalization without the need for retraining, validating its potential for collaborative, energy-efficient management in next-generation multi-operator networks.