Analysis of MAC schemes for cognitive radio network: Perfect and imperfect learning modelling

In cognitive radio networks (CRNs), secondary users (SUs) opportunistically utilize the spectrum without harmful interruption to the primary users (PUs). We consider a distributed learning algorithm arisen in the context of CRNs. There are multiple distributed SUs searching for idle channels temporarily unused by the primary network. Optimal MAC is also analyzed with multiple channels and multiple secondary users (SUs). MAC solution for cognitive radio can also be implemented by the perfect and imperfect learning scheme. Perfect or partially observed Markov decision process (POMDP) framework is suitable when the PU´s transition probabilities are known or partially known to the SUs. But in the CRNs, SUs are not exactly aware about the presence of the primary user´s transmission in the particular spectrum band. So, imperfect learning model is also proposed for these type of unknown transition probability of primary users. Both perfect and imperfect learning is considered with and without the acknowledgement by the receiver. This MAC technique optimizes SU access policy while protecting PU performance. This paper also try to design optimal decentralized MAC scheme under a constraint on the probability of primary collisions.