Backhaul aware joint uplink and downlink user association for delay-power trade-offs in HetNets with hybrid energy sources

In cellular networks, conventional user association algorithms are solely based on downlink (DL) performance, which may lead to inefficient transmission power and high interference in the uplink (UL) transmission. In addition, the backhaul data rate constraint has been neglected by the majority of the existing user association algorithms. However, the backhaul constraint has become more severe in heterogeneous networks (HetNets) where small cells are densely deployed to meet the skyrocketing data traffic demand. In this paper, we propose an optimal backhaul-aware joint UL and DL user association for delay-power trade-offs in HetNets with hybrid energy sources. In the considered HetNets, all the base stations are assumed to be powered by a combination of power grid and renewable energy sources, in order to achieve both uninterrupted and green communications. Taking both UL and DL transmissions into consideration, the proposed user association algorithm aims to improve network quality of service by minimising the sum of UL and DL average traffic delay, as well as to reduce the overall UL power consumption of users and DL on-grid power consumption by maximising the utilisation of green power harvested from renewable energy sources. To this end, a convex optimisation problem is formulated to minimise the weighted sum of cost of average traffic delay and cost of power consumption. We have proved that the proposed user association algorithm converges to the global optimum, which enables a flexible trade-off between average traffic delay and power consumption. Simulation results validate the effectiveness of the proposed algorithm in adapting the traffic loads among base stations along with the distribution of green power and the backhaul data rate constraint. Simulation results also demonstrate that the proposed user association algorithm achieves prominent improvement in UL average traffic delay reduction and effectively reduces both the DL on-grid power consumption and overall UL power consumption of users, with limited sacrifice on DL average traffic delay, compared with the user association algorithm only based on DL performance. Copyright © 2015 John Wiley & Sons, Ltd.