Blind cognitive-MAC: A Bayesian game theoretic perspective

The protocol design of the medium access control (MAC) is critical for the newly promising paradigm of cognitive networks (CNs). The current studies have generally assumed completeness of the context information, or there is a central control node, wherein a common control channel is often assumed. This is not practical for the distributed CNs, and always leads to enormous computation and high complexity of implementation of the designed cognitive MAC. To avoid these problems, from a Bayesian game theoretical perspective, a blind cognitive MAC is investigated in this paper with the objective of maximizing data throughput of the individual selfish and rational secondary user (SU). This is different from the literature, which employed the Nash Game and formulated as the Bargaining Game. Entirely from their own perspective, the SUs sense, understand, estimate and learning/resonsing the cognitive context, mainly regarding to the dynamic behaviors of primary users (PUs). The proposed MAC is with the self-awareness of the SU, so we call it as the blind cognitive-MAC (BC-MAC). Numerical results demonstrate the existence of the Bayesian equilibrium solution of BC-MAC, the convergence of the Bayesian learning/reasoning approach (BLA), and the performance of the proposed BC-MAC protocol. This paper is also motivated by the recent work, but we investigate how to use Bayesian games to investigate the cognitive MAC design. The main contributions of our work are summarized as: Utilize Bayesian game to model the cognitive MAC problem, which focus on the incomplete information case and avoid to use the common control channel, as well as take the MAC predictability into consideration. Propose the prediction approach (BLA) for exactly context awareness of S_u during the decision making process. Present the BC-MAC protocol, and verify the existence of optimal strategy, convergence of BLA algorithm and the performance of presented BC-MAC protocol.