A new distributed algorithm for the allocation of cognitive radio sensing times

In cognitive radio systems, the sparse assigned frequency bands are opened to secondary users. We consider the sensing of the frequency spectrum for cognitive radios, based on energy detection. It has been shown that the sensing reliability can be improved in the presence of fading by using several cooperating cognitive radios that exchange their individual sensing information to a coordinator node. The coordinator node combines the received information in order to make a decision about the primary network presence. The primary network sensing times used by the cognitive nodes should be chosen high enough to ensure the correct detection of the primary emitter but low enough so that the nodes still have enough time to communicate. In this paper, we propose a decentralized Q-Learning algorithm to allocate the sensing times of the cognitive radios in a way that maximizes the throughputs of the radios. A rigorous proof of the convergence of the proposed algorithm is provided. Numerical results show that it converges faster to the same throughputs as those obtained with a reference algorithm.