Joint User-Target Pairing, Power Control, and Beamforming for NOMA-Aided ISAC Networks

Abstract

Integrated sensing and communication (ISAC) emerges as a pivotal solution for augmenting spectrum efficiency and fostering synergies between sensing and communication functionalities. However, ISAC efficacy grapples with inter-functionality interference, which can be efficiently managed by non-orthogonal multiple access (NOMA) schemes. Accordingly, this paper unveils multi-armed bandit (MAB)-based approaches, interplaying between communication throughput and radar estimation metrics. Our optimization challenge seamlessly transitions from a multi-objective problem to a weighted sum single-objective problem, exploiting two MAB variants-the decaying $\epsilon $ -greedy and the upper confidence bound. Both algorithms manage interference in NOMA-ISAC by jointly designing power allocation and pairing of communication users and radar targets. To improve convergence rates, a multi-MAB approach is proposed, dividing the network into partitions, each managed by a dedicated single MAB agent. We also propose three beamforming methods; 1) zero-forcing beamforming based decoding method, 2) two-step MAB approach, commencing with the ZF-BF and succeeding with a subsequent MAB phase to bolster beamforming efficacy, and 3) beam-sweeping-based technique for scenarios with CSI absence, utilizing the discrete Fourier transform (DFT) codebook. Numerical results validate the efficacy of the proposed algorithms, outperforming conventional techniques by an average of 65%, closely approaching the exhaustive search by only 2% with approximately 95% less computational complexity.