Channel sensing order for cognitive radio networks using reinforcement learning

This work investigates the problem of channel sensing order used by a cognitive multichannel network, where each user is able to perform primary user detection on only one channel at a time. The sensing order indicates the sequence of channels sensed by the secondary users when searching for an available channel. When using an optimal sensing order, the secondary user can find faster a free channel with high quality. Brute-force algorithms may be used to find the optimal sensing order. However, this approach requires great computational effort. Even in scenarios where the secondary user knows the probability of each channel being available, the sensing order where the most available channels are sensed first is not ideal when using adaptive modulation. Therefore, we propose an approach using reinforcement learning to search dynamically for the optimal sensing order. Through simulations, we evaluated our proposal and compared its performance with other mechanisms, and the results obtained are close to the optimal value provided by the brute-force and superior to the other mechanisms in most of the scenarios.