Comparison of Energy Aware and Cost Subsidy Multi Armed Bandit Solutions for Supreme Channel Election in Hybrid Band Systems

Abstract

Recent wireless communication systems, such as device-to-device (D2D) communications and internet of things (IoT), etc., support hybrid band frequencies to sustain the user demands in B5G/6G systems throughout switching between bands and avoid connection loss. This paper compares two online learning solutions for optimal band/channel assignment in hybrid radio frequency (WiFi and WiGig) and visible light communication (RF/VLC) wireless systems. In such scenarios, the multi-band source/transmitter (S/Tx) has no prior knowledge about distinct channel characteristics, including their transmission rates and consumed energy. Therefore, to extend its limited battery, the S/Tx has to target the best arm/band with the least possible consumed power. Hence, we compare two Multi Armed Bandit (MAB)-based solutions, which are costsubsidy MABs (CSMABs), where the S/Tx sacrifices with the highest reward in order to select the lowest cost arm/operating frequency and energy-aware MABs (EAMABs) where the cost term is amended only to the exploration term. In both methods, the S/Tx targets to maximize his cumulative payoff (transmission rate) and minimize his cost (battery expenditure due to the operating band/frequency). Numerical simulations indicate that proposed CS-MAB schemes outperform purely explored MABs via Thompson sampling (TS), upper Confidence bound (UCB), and benchmark multi-band election (MBE) approaches, correspondingly in terms of transmission rates and energy efficiency.