A Secured Energy Saving With Federated Assisted Modified Actor-Critic Framework for 6G Networks

Ultra-dense base station deployments must be operated in a way that allows the network's energy consumption to be adapted to the spatio-temporal traffic dynamics, thereby minimizing overall energy consumption. To achieve this goal, we leverage two artificial intelligence algorithms—federated learning and actor-critic—to develop a proactive and intelligent cell switching framework. This framework can learn the operating policy of small base stations in an ultra-dense heterogeneous network, resulting in maximum energy savings while respecting quality of service (QoS) constraints. Additionally, the use of federated learning enhances the security of the system as only the model parameters are shared among the entities, rather than the raw and potentially sensitive data. In other words, in the event of a data breach or eavesdropping, only the parameters of the developed artificial intelligence model would be compromised. This approach ensures that the network remains secure against attacks targeting confidential and sensitive data. The performance evaluation reveals that the proposed framework can achieve an energy saving that is about 77% more than that of the state-of-the-art solutions while respecting the QoS constraints of the network.