Spectrum Sensing Across Multiple Service Providers: A Discounted Thompson Sampling Method

Abstract

Dynamic spectrum access is a potential solution to the long-standing radio spectrum scarcity and usage inefficiency issue, for which spectrum sensing is one of the key challenges. In this letter, we investigate the spectrum sensing order problem in the scenario that the secondary user equipped with limited sensing capability can get access to the idle spectrum across multiple network service providers opportunistically by spectrum sensing, where the probability of the spectrum being idle varies at temporal scale and is not available for the users. We employ an online learning method, referred to as discounted Thompson sampling, to address the formulated optimization task, which can track the changes of the probability of the spectrum availability and yields more spectrum access opportunities compared to other methods.