Channel sensing order for distributed cognitive networks with multi-user and multi-channel

This work investigates the problem of channel sensing order in distributed cognitive networks with multiuser, in which the set of active secondary users is randomly changing in each time slot. We model this problem into a non-cooperative game and define a generalized interference metric. The optimization objective is to minimize the generalized interference metric. To cope with this problem, we propose Q learning algorithm and no-regret learning algorithm. Through simulations we evaluate our proposal and compare its throughput performance with random selection algorithm, the results obtained by Q learning algorithm are close to no-regret learning algorithm and superior to random selection algorithm.