Decentralized cognitive MAC protocol based on SARSA

In this paper, we propose a decentralized cognitive MAC protocol based on SARSA that allow secondary users to independently seek opportunistic spectrums without exchanging any control information among them. Due to sensing equipment constrains, dynamic spectrum access has become a major problem faced by the cognitive radio in an uncertain environment. Partially Observable Markov Decision Process (POMDP) is considered to be an effective way to solve this problem. Under a POMDP framework, the MAC protocol we proposed, which has a self-learning and online learning ability, does not require the channel state transition probability of the primary network, and can take full advantage of the past sensing and access experiences to optimize current channel access. The simulation results show that the proposed approach can effectively reduce the conflict probability between the second user (SU) and the primary user (PU), as well as SU and SU, and improve system throughput and spectrum efficiency.