Optimal resource preemption for aperiodic URLLC traffic in 5G Networks

In this paper, we study DownLink Preemption (DLP), a feature enabling dynamic scheduling for Ultra Reliable Low Latency Communications (URLLC) in the presence of ongoing enhanced Mobile Broad Band (eMBB) flows. We design a DLP technique with the objective of maximizing the admission of URLLC packets while minimizing the impact of preemption on the eMBB throughput. We propose two different approaches to solve the DLP problem. A first, the Offline Preemption Approach (OPA), is formulated as a multi-objective Integer Linear Program (ILP) and considers a perfect knowledge of traffic dynamics. We further propose the Impact-Aware Preemption Approach (IAPA), an algorithm solving the DLP problem and performing joint admission and preemption decisions on the fly. We conduct extensive simulations using a system level simulator in realistic 5G Network settings. Our numerical results demonstrate the efficiency of IAPA in guaranteeing a close-to-optimum performance.