Collision reduction in cognitive radio using multichannel 1-persistent CSMA combined with reinforcement learning

In this paper a novel multiple access scheme, M-CSMA-RL, is proposed for secondary users which combines multichannel 1-persistent CSMA and reinforcement learning. The scheme effectively reduces the probability of packet collisions among primary and secondary users sharing common spectrum. Compared with multichannel CSMA without learning, the throughput and packet loss of M-CSMA-RL shows a significant improvement in a distributed cognitive radio scenario in situations where primary users operate with TDMA/FDMA. The results show how the M-CSMA-RL scheme improves both primary and secondary user's throughput at various offered traffic levels and with different ratios of primary and secondary user offered traffic.