Minimizing the Queue-Length-Bound Violation Probability for URLLC: A Cross Layer Approach

Shaoling Hu,Wei Chen
DOI: https://doi.org/10.1109/globecom42002.2020.9322353
2020-01-01
Abstract:Most recently, Ultra-Reliable Low-Latency Communication (URLLC) has attracted much attention due to its potential application in Factory Automation (FA), Vehicle-to-everything (V2X), telesurgery, etc. One of the key performance metrics for URLLC is the Queue-Length-Bound (QLB) violation probability. However, how to minimize the QLB violation probability remains open over time-varying channels because it relies on challenging cross-layer designs. In this paper, we formulate a Constrained Markov Decision Process (CMDP) framework to minimize this QLB violation probability while keeping the average transmission power low. To achieve a feasible solution, we then relax this CMDP problem into an unconstrained Markov Decision Process (MDP) and apply the value iteration algorithm. By this means, we acquire the optimal stationary deterministic policy. It is further shown that a threshold based policy can strike the optimal QLB violation probability and power tradeoff. Finally, a Linear Programming (LP) is adopted to solve this CMDP given a certain power constraint. We thus acquire the optimal probabilistic policy and its tradeoff between the QLB violation probability and the transmission power.
What problem does this paper attempt to address?