A New Opportunistic Spectrum Access Approach in Unslotted Primary Systems

To solve the channel selection issue in opportunistic spectrum access, a new Q-Learning based algorithm for channel selection is proposed. The algorithm can lead secondary user to select channels with maximum cumulative reward, and maximize secondary user throughput. A Boltzmann learning rule is adopted to achieve well tradeoff between channel exploration and exploitation. From the simulation results, compared with random selection algorithm, the algorithm does not require prior knowledge or prediction models of the channel environment, yet can still select the optimal channel adaptively, improve the secondary user capability and attain to the convergence in short time.