Capturing Rare Network Conditions to Dimension Resources for Ultra-Reliable Communication

Dimensioning for ultra-reliable communication is challenging because it requires massive amounts of data to capture rare network conditions, which is time-consuming, or even impractical, depending on the time it takes to collect the data. This letter proposes an extreme value theory-based framework to reduce the demand for data while quantifying the bandwidth needed to meet ultra-reliability requirements. Our proposed approach is shown to use several orders of magnitude less data than what would be conventionally expected when tested in two 3GPP network scenarios. It also consistently converges in ultra-reliability regimes using fewer data samples than comparable state-of-the-art methods.