Cost-Aware Bandits for Efficient Channel Selection in Hybrid Band Networks

Recently, hybrid band communications have received much attention to fulfil the exponentially growing user demands in next-generation communication networks. Still, determining best communicate over is a challenging issue, especially dynamic channel conditions multi-band wireless systems. In this paper, we manipulate practical online-learning-based solution for band/channel selection radio frequency and visible light (RF/VLC) The difficulty formulated as multi-armed bandit (MAB) with cost subsidy, which learner (transmitter) endeavors not only increase his total reward (throughput) but also reduce (energy consumption). Consequently, propose two (HBS) algorithms, named subsidy upper confidence bound (CSUCB-HBS) Thompson sampling (CSTS-HBS), efficiently handle problem obtain high throughput low energy consumption. Extensive simulations confirm that CSTS-/CSUCB-HBS outperform naive TS/UCB heuristic HBS approaches regarding consumption, efficiency, throughput, convergence speed.