Energy Efficiency and Coverage Probability

The non-Poisson characteristics of real-world cellulardeployments complicate the modeling and analysis of keymetrics namely coverage probability (Pcov) and energy efficiency(EE). This letter investigates the use of empirical data fromsystem-level simulations to approximate Pcov and EE as posynomialfunctions of base station (BS) transmit power and deploymentdensity using convex regression. These posynomial-basedmodels are then used to formulate an EE optimization problemunder a coverage constraint as a geometric programming (GP)problem. The optimal solution to this GP is proposed based onduality. The effectiveness of this approach is validated throughnumerical results across various network scenarios, includingnon-Poisson BS topologies and varying LoS ranges.