On Optimality of Myopic Policy in Opportunistic Spectrum Access: The Case of Sensing Multiple Channels and Accessing One Channel

Recent works have developed a simple and robust myopic sensing policy for multi-channel opportunistic communication systems where a secondary user (SU) can access one of N i.i.d. Markov channels. The optimality of the myopic sensing policy maximizing the SU's expected accumulative reward is established under certain conditions on channel parameters. This paper studies the scenario where the SU, equipped with one radio, can sense k channels but access only one channel each time slot. The objective of the SU is to maximize its throughput over the time horizon T. By characterizing the myopic sensing policy in this context, we establish analytically its optimality for a subset of specific system settings, notably the cases k=2, T=2 and k=N-1 with arbitrary T. In the more generic case, we construct counterexamples to show that the myopic sensing policy is not optimal.