A Reinforcement learning-based cognitive MAC protocol

A Multi-Channel Cognitive MAC Protocol for adhoc cognitive networks that uses a distributed learning reinforcement scheme is proposed in this paper. The proposed protocol learns the Primary User (PU) traffic characteristics and then selects the best channel to transmit. The scheme, whichaddresses overlay cognitive networks,avoids collision with the PU nodes and manages toexceed the performance of the less adaptive statistical channel selection schemesin normal and especially bursty traffic environments. The simulation analysis results have shown that the performance of our proposed scheme outperforms that of the CREAM-MAC scheme.