Distributed, Inference-Based, Energy Efficient User Association with Convergence Guarantees

This work establishes a relaxed formulation for the problem of user association with energy efficiency (EE) in the downlink of heterogeneous networks. A distributed solver is offered, based on belief propagation (BP)-based message passing between (micro or macro) base stations. In addition, correctness and convergence guarantees are provided for the employed loopy BP and computation load is analyzed and reduced (with careful precomputations). Numerical results indicate that the proposed approach offers higher EE geometric mean, while not sacrificing spectral efficiency, compared to prior art.