Abstract
The millimeter-waves (mmWaves) features promote its soon employment in device to device (D2D) communications. In D2D, neighbor discovery and selection (NDS) problem is a critical one due to the tradeoff between exploring more devices for the best choice and the expected beamforming training (BT) overhead. In this paper, mmWave D2D neighbor discovery and selection (NDS) problem is modeled as a budget-constrained multiarmed bandit (MAB). Specifically, an energy constrained minimax optimal stochastic strategy (E-MOSS) algorithm is proposed, which reflects the real network scenario by counting the remaining battery levels of the neighboring devices. Simulation results prove the efficiency of the proposed E-MOSS algorithm over the traditional NDS arrangements regards network lifetime, convergence rate, energy performance, and average throughput. Index Terms-MOSS, mmWave, D2D, Multiarmed Bandit (MAB).