A Multi-Armed Bandit Approach for User-Target Pairing in NOMA-Aided ISAC

In this paper, we propose a robust interference management approach for the integrated sensing and communication (ISAC) system that employs non-orthogonal multiple access (NOMA) for multiplexing. Our proposed approach effectively addresses interference challenges by optimizing the pairing of communication users (CUs) and radar targets (RTs) while simultaneously designing receiving beamformers. These optimizations aim to maximize the combined utility of communication rates and the radar estimation information rate (REIR), inherently constituting a challenging non-convex combinatorial problem. To tackle this intricate problem, we employ the upper confidence bound (UCB) algorithm, a powerful online learning technique rooted in multi-armed bandit (MAB) theory. Along with UCB, we harness zeroforcing beamforming to optimize the receiving beamformer. The numerical results underscore the importance of CU-RT pairing, with a $65 \%$ average performance improvement over traditional NOMA-ISAC and OMA-ISAC, close to the exhaustive search performance by only $2 \%$. It also substantially reduces complexity, with about $90 \%$ less computational complexity than exhaustive search.