Low Latency mmWave Backhaul via Traffic Dispersion

This paper focuses on reducing the beamtraining delay and end-to-end latency using »traffic dispersion» in a multi-hop millimeter-wave (mmWave) backhaul link. Traffic dispersion reduces the latency by creating multiple parallel paths between a base station (BS) and the anchored base station (ABS) connected to a core network, to increase the data rate. The beamtraining delay is a major source of end-to-end latency and is particularly crucial in case of outdoor deployment of the mmWave backhaul link, suffering from frequent beam outage due to beam misalignment caused by wind or ground vibrations, thereby lowering its coherence time. We formulate an optimization problem to minimize the beamtraining delay by jointly optimizing the power allocation and beamwidths in paths between the BS and intermediate relay units (RU) and that between RUs and ABS, subject to an end-to-end latency requirement. The problem is mix-integer and non-convex in nature which is challenging to solve in general. Thus, we address the problem using the iterative Practical Swarm Optimization (PSO) in conjunction with penalty function method and primal decomposition, where the transmit power and the beamwidths are iteratively updated. Numerical results show the robustness of the proposed algorithm in reducing the beamtraining delay and end-to-end latency for mmWave backhaul link with a varying coherence time.