Self-Organizing Spectrum Breathing and User Association for Load Balancing in Wireless Networks

In this paper, we develop a self-organizing mechanism for spectrum breathing and user association in cellular networks employing frequency reuse patterns. Specifically, our focus is on flow-level cell load balancing under spatially inhomogeneous traffic distributions. Our work adaptively changes the spectrum bandwidth of each base station (BS) so that spectrums of BSs breathe in and out in order to balance the loads of BSs. Spectrum breathing is further combined with delay-optimal user association for better load balancing. Our problem is challenging because the problem is not a convex optimization. To tackle the difficulty, we decouple spectrum breathing and user association and propose an iterative algorithm that always converges to a fixed point, which is possibly an optimal solution. We show that spectrum breathing dominates a family of α-optimal user association in cell load balancing. Surprisingly, the flow-level delay performance under spectrum breathing gets even better as spatial traffic distribution becomes unbalanced, which is not the case of α-optimal user association. Our extensive simulations confirm that spectrum breathing significantly improves the system performances: decreasing the delay more than 10 times or increasing the admittable traffic load by more than 125%. Furthermore, spectrum breathing outperforms full frequency reuse when spatial traffic distribution is inhomogeneous.