Joint Resource Allocation for Underwater Acoustic Communications Using Self-Estimation and Exploration Regulation MAB

초록

This paper proposes a joint channel and power resource allocation scheme based on the self-estimation and exploration regulation multi-armed bandit (SEER-MAB) framework for underwater acoustic (UWA) communication systems. The proposed method is built on an adversarial MAB and incorporates both real reward-based weight updates and self-estimated reward refinement guided by channel variability. Combining these mechanisms, SEER-MAB effectively balances exploration and exploitation while providing robustness to time-varying underwater channel conditions. Numerical simulation results demonstrate that SEER-MAB outperforms the conventional Exp3 algorithm, achieving improvements of 29.2% in utility and 47.9% in cumulative regret. © 2025 IEEE.

키워드