LAUSR

Ultra-dense heterogeneous networks (UDHNs) have been widely regarded as a promising solution to enhance the reuse efficiency of spatial frequency and thus improve the overall network performance. In such networks, a load balancing problem caused by coexisting macrocells and small cells should be treated seriously, which means some effective user associations are essential for load balancing. To this end, we design two types of offloading (load balancing) schemes for UDHNs to maximize a logarithmic utility of long-term rates. To guarantee the load balancing gain, a frequency partitioning scheme is designed to degrade the cross-plane interference, and a time partitioning strategy is developed to eliminate the strong interference received by some users offloaded from macrocells. In these offloading schemes, the main difference between them is whether a time partitioning factor needs to be optimized. As for the problems formulated in these schemes, we design a distributed algorithm by utilizing dual decomposition and develop a centralized algorithm with a two-layer iteration. Then, we give some detailed convergence and complexity analyses for them. Numerical results show that the proposed schemes yield some significant performance gains relative to some traditional ones, and the centralized algorithm often achieves a better association performance than a distributed one since the former almost always tries to optimize a time partitioning factor.