Delay Optimal Beamformer Design for Cache-enabled Wireless Backhaul Networks

We consider the downlink transmissions in a twotier ultra-dense wireless network, whereby data is delivered from the base station (BS) to the cache-enabled access points (APs) through wireless backhaul links using multicast beamforming technique, and then the users are served by the APs using joint beamforming technique in the access links. To reduce the transmission delay incurred in the backhaul links, the requested data may be cached in some APs in advance during the off-peak time, and thus the users can download the data from the cached APs directly. However, under some circumstances, it is beneficial to select some un-cached APs to the serving set as long as the reduced delay in the access link is more than the delay penalty in the backhaul links. We formulate this problem as a delay optimal beamformer design (DOBD) problem, which attempts to minimize the transmission delay experienced in the backhaul and access link transmission stages. This problem is nontrivial due to the non-convexity and coupling between backhaul and access links, which is transformed into a tractable form using the semi-definite relaxation (SDR) and sequential convex approximation (SCA) schemes, and thus can be solved efficiently using an iterative algorithm. Simulation results demonstrate that the proposed scheme can converge to the stationary point quickly under different network settings, and the overall transmission delay can be reduced significantly comparing with the conventional schemes.