PHY-Aware MCS Adaptation for NR V2X Mode 2 with ML-Driven Resource Allocation

Abstract This paper proposes a unified framework to enhance NR V2X sidelink Mode 2 by combining PHY-aware MCS adaptation with reinforcement learning (RL)-based resource allocation. In out-of-coverage scenarios, vehicles select MCS based on real-time SINR using thresholds from the UWICORE dataset, while an RL agent optimizes semi-persistent scheduling to avoid collisions. Analytical models for packet delivery, collision probability, and spectral efficiency are developed and validated using WiLabV2X simulations. Results demonstrate improved PDR, reduced collisions, and higher energy and spectral efficiency over standard SPS, particularly in dense traffic, aligning with 3GPP TR 37.885 for safety-critical use.