Delay Optimal Beamformer Design For Cache-enabled Wireless Backhaul Networks

Traffic,user numbers and user density has increased swiftly in decent years.To support bigger network traffic,more user numbers and higher user density with better service quality,we need to research on methods to improve performance of wireless network.However,conventional solutions like finding more frequency band,increasing station power has been difficult or too expensive.Ultra dense network,UDN,is considered as a new solution to improve performance for 5G networks,which deploy many low power access points in high density within their serving area to serve more users with lower AP power consumption.However,dense AP in UDNs will cause bigger interference among users.As a result,coordinate beamforming is introduced in access links.On the other hand,there is another challenge in UDN networks,APs are deployed in locations near users,where wired backhaul links cannot be deployed.As a result,we must introduce wireless backhaul technology in UDN.The wireless connection is easier to be influenced by inter-user interference,so the wireless backhaul transmission becomes the bottleneck of the network performance.As a solution,cache is introduced at APs,enabling to draw popular data from BS before the user requesting them then eliminate the backhaul delay when the data need to be transmitted.However,under some circumstances,it is beneficial to select some un-cached APs into the serving AP set of a user,because the newly introduced backhaul delay penalty can be covered by the access delay reduction accounting for an additional coordinating AP for the user.Therefore,we consider a delay optimal beamformer design problem,which minimizes the overall delay in the two stages of access and backhaul.We formulate the DOBD problem under both the non-fragment and fragment caching strategies.The DOBD problem is a nontrivial problem caused by its non-convexity of backhaul and access constraints.We propose to transform the original problems into a tractable convex from using the semi-definite relaxation and sequential convex approximation methods.Then we design an iterative algorithm under SCA scheme to solve the transformed problem efficiently.We then show some simulation results to demonstrate that the proposed algorithm can converge quickly under different scenarios,and reduce the overall transmission delay comparing with the conventional schemes under both non-fragment and fragment strategies.