Power and Rate Adaptation for URLLC With Statistical Channel Knowledge and HARQ

This letter investigates a point-to-point ultra-reliable low latency communication (URLLC) transmission with statistical channel knowledge. We consider different hybrid automatic repeat request (HARQ) schemes and investigate the signal-to-noise ratio (SNR) feedback from failed packets to improve transmission efficiency. The problem is formulated as a long-term power minimization problem under URLLC requirement. A deep reinforcement learning (DRL) agent, employing proximal policy optimization (PPO), is used to control transmit power and the coding rate dynamically to solve the formulated problem. Simulation results demonstrate HARQ strategies, SNR feedback, and PPO algorithm can bring significant gains and reveal the impact of reliability and latency.