Allocation of licensed and unlicensed spectrum in heterogeneous networks

In order to support fast growing mobile data traffic, wireless operators have started to use multiple radio access technologies (RATs) over multiple licensed and unlicensed frequency bands. A widely used method is WiFi off-loading of cellular network traffic. A recent proposal is side-by-side deployment of Long Term Evolution (LTE) in licensed spectrum and LTE in unlicensed spectrum (LTE-U). This paper studies the spectrum allocation problem in such heterogeneous networks (HetNets). For practical reasons, the allocation is conceived to be on a relatively slow timescale. A queueing model is introduced for the unlicensed band to capture its lower spectral efficiency, reliability, and additional delay due to contention and/or listen-before-talk requirements. Under mild assumptions, the spectrum allocation problem is formulated as a bi-convex optimization problem. Solving this problem gives an effective and computationally efficient solution for both user association and spectrum allocation over multiple RATs. Simulation results show that in the heavy-traffic regime, the proposed scheme significantly outperforms both orthogonal and full-frequency-reuse allocations. In addition, the solution to the optimization problem matches the intuition that users with relatively higher traffic demand are mostly assigned to the licensed spectrum, while those with lower traffic demand and less exogenous interference from the unlicensed band are assigned to the unlicensed spectrum.