Epoch-Based Channel Selection Algorithm for Device-to-Device Communication Using Multi-armed Bandit with Random Rewards and Adaptive Randomization

Abstract

Abstract In wireless networks, device-to-device communication continues to rise significantly. Due to the high performance and bandwidth requirements of an increasing number of users, network capacity would be substantially limited by spectrum availability soon. Besides, due to the fixed assignment policy, the present network scenario is experiencing a vast spectrum shortage problem, so a substantial spectrum remains unused. Opportunistic spectrum access (OSA) is a method used to increase the efficiency of the spectrum. Thus, to facilitate the device-to-device communication by OSA, a truly distributed channel selection algorithm is proposed. Besides, the mean availability of different channels is unknown. Thus, identification and selection of the best channel become more difficult. Moreover, it is considered that no information exchange takes place between different users in order to access the shared spectrum opportunistically. Users must transmit on those channels most likely to be vacant to achieve the optimum network throughput. Hence, it is essential to estimate the mean availability of different channels at the individual node. An algorithm based on the multi-armed bandit (MAB) is proposed, termed epoch-based channel selection (EBCS). The results indicate that the regret of the proposed algorithm is logarithmic. The collisions are less than 2% of total time slots.KeywordsMulti-armed banditsDecentralized networkDistributed algorithmsChannel selection policyOpportunistic spectrum access