Efficient Service Differentiation and Energy Management in Hybrid WiFi/LiFi Networks

In this paper, we propose a predict-and-optimize algorithm for hybrid WiFi/LiFi networks to enable service differentiation while maximizing energy efficiency (EE). The framework dynamically predicts the appropriate network slice for each user based on historical key performance indicators (KPIs), using a deep learning model trained with the resilient backpropagation algorithm. Subsequently, the algorithm optimizes EE as a function of precoding vectors for the predicted slices using sequential convex approximation and the inner approximation method. Novel approximations are introduced to convert non-convex objectives and constraints into convex forms, enabling an iterative algorithm to compute sub-optimal solutions. The EE formulation complies with end-to-end latency requirements and addresses hybrid-specific constraints, such as LED input signal limitations, data rate thresholds, and power budgets. Slice-specific quality of service (QoS) demands are incorporated by modeling enhanced mobile broadband (eMBB) with infinite blocklength capacity and using finite blocklength approximations for ultra-reliable low-latency communication (URLLC) and massive machine-type communications (mMTC), ensuring realistic reliability and latency treatment. Simulation results demonstrate that the proposed algorithm achieves up to 37% EE improvement over Zero-Forcing (ZF) precoding. It converges within 10–20 iterations and exhibits robust scalability with increasing numbers of access points, validating its effectiveness for next-generation heterogeneous wireless networks.