An upper bound on the throughput for myopic policy in multi-channel opportunistic access

Abstract

We study myopic sensing policy for an opportunistic communication system in which the states of channels evolve as independent Markov chains. A Secondary User (SU) takes one channel to sense and access in each time slot by myopic sensing policy. In the myopic policy, SU ignores the impact of the current action on the future decisions and tries to maximize the expected immediate reward. We propose an upper bound on the throughput achieved by the myopic sensing policy for general case in which channels have different Markov models. In particular, the proposed bound is compared with the Zhao et. al's upper bound (UZhao) for the channels with identical models. We prove that there are conditions in which the proposed bound is tighter than UZhao bound. Besides, the throughput obtained from the simulation is compared with the proposed bound. The results demonstrate that the bound is tight and close to the simulation results.