A Data-Driven Cost-Effective Session-Oriented Cognitive Radio Transmission Scheme Under Spectrum Uncertainty

Abstract

Due to the emerging Internet of Things (IoT) services, the spectrum shortage problem becomes more and more serious. To tackle this challenge, many research works have been conducted to employ the cognitive radio technology to exploit under-utilized spectrums for IoT services. However, the operation of a cognitive radio transmission system is usually time-energy-consuming due to the requirement on the wideband sensing and spectrum switching, which might be hardly supportable by the light-weighted IoT devices. In this paper, we propose a data-driven cost-effective session-oriented cognitive radio transmission scheme, where the bands are directly selected based on the historical data and a “transmit-wait-transmit” mode is employed to reduce the cost. For the spectrum selection, we first attempt to determine the bands with minimal total bandwidth that could make the session accomplished with certain confidence level by modeling the available duration of a band within the session period as a random variable. Then, from the historical data, we develop a distributionally robust approach, where Kullback-Leibler divergence is used to capture the distributional ambiguity. Finally, based on the real data we collected using USRP-2922, we evaluate the effectiveness of our proposed scheme.