AI and the World's Hard Problems · Theoretical computer science · arXiv · 2026-10-26
Digital-Twin-Referenced Belief-Guided Q-Learning for Jammer Identification and Beam-Power Control in a mmWave Model
Joint jammer identification and beam-power control in 6G mmWave networks is the problem addressed here. The authors use a simulation-based digital-twin reference, against which the actual network state is assessed, and a belief-guided Q-learning (BGQL) method for a site-specific 28-GHz network model. Across 30 paired Monte Carlo runs, compared with SARSA, Greedy and Random, BGQL achieved link-rate scores of 437.4–564.2 Mbit/s and localization recall of 100% for one jammer and 78–81% for two jammers, though each run took 80–130 s versus 5–14 s for SARSA. The authors present this as a simulation proof of concept.