Reinforcement Learning for Delay Sensitive Uplink Outer-Loop Link Adaptation

Because modulation and coding scheme (MCS) selection and scheduling are always made upfront, link adaptation (LA) needs to balance between reliability and spectral efficiency with outdated signal-to-interference-plus-noise ratio (SINR) information. Moreover, due to demands of new 5G use cases and rather limited user equipment (UE) transmission power budget, required high reliability and low latency transmissions are becoming even more challenging, especially in challenging propagation environments. In this paper, online reinforcement learning (RL) is harnessed to tackle this LA related problem. In particular, we propose practical Q-learning-based outer-loop link adaptation (OLLA) algorithm that aims at using minimal amount of radio resources for delivering packets within packet delay budget (PDB). By striving for low radio resource usage also multi-agent RL related greediness problems can be mitigated. Realistic system level simulations confirm that the proposed algorithm outperforms traditional OLLA when high reliability and low delays are required. It is also shown that the challenging requirements, set by extended reality (XR) uplink traffic, can be reached.