Multi-Channel Opportunistic Access Based on Primary ARQ Messages Overhearing

Optimality of Myopic Policy for Restless Multiarmed Bandit with Imperfect Observation

“…Specifically, a simple myopic policy, also called greedy policy, is developed in [7] which yields a factor 2 approximation of the optimal policy for a subclass of scenarios referred to as Monotone MAB. The second thrust is to establish sufficient conditions to guarantee the optimality of the myopic policy in some specific instances of restless bandit scenarios, particularly in the context of opportunistic communications [10][11][12][13][14][17][18][19].…”

“…ones, and focused on a class of so-called regular functions, and derived closed-form sufficient conditions to guarantee the optimality of myopic sensing policy. The authors [17] studied the myopic channel probing policy for the similar scenario proposed, but only established its optimality in the particular case of probing one channel (M = 1) each time. In our previous work [18], we established the optimality of myopic policy for the case of probing N − 1 of N channels each time and analyzed the performance of the myopic probing policy by domination theory, and further in [19] studied the generic case of arbitrary M and derived more strong conditions on the optimality by dropping one of the non-trivial conditions of [17].…”

Myopic policy for opportunistic access in cognitive radio networks by exploiting primary user feedbacks

“…For the similar scenario considered in this paper, the authors of [20] established the optimality of the myopic policy for the case of probing one channel ( k = 1) each time. In our previous work [21], we established the optimality of myopic policy for the case of probing N − 1 of N channels and analysed its performance by domination theory, and then extended the optimality to probe multiple channels for positively correlated homogeneous channels (PCOC) [1].…”

“…In our previous work [21], we established the optimality of myopic policy for the case of probing N − 1 of N channels and analysed its performance by domination theory, and then extended the optimality to probe multiple channels for positively correlated homogeneous channels (PCOC) [1]. Compared with the most relevant literature [1, 20, 21], in this paper we derive the sufficient conditions to guarantee the optimality of the myopic policy for four different cases: negatively correlated homogeneous channels (NCOC), heterogeneous channels (EC), positively correlated EC (PCEC) and negatively correlated EC (NCEC). Specifically, for NCOC, the reverse structure of belief vector preserves three critical exchange operations in branch and bound process concerning the derivation of the optimality of myopic policy.…”