Blind cyclostationary feature detection based spectrum sensing for autonomous self-learning cognitive radios

In this paper, we present an autonomous cognitive radio (CR) architecture that incorporates the main features of cognition. This model, referred to as the Radiobot, is capable of self-learning and self-reconfiguration to match its RF environment. The proposed CR architecture assumes a joint blind energy and cyclostationary detection methods to classify the communication systems in its vicinity, without any prior knowledge of the sensed signals. We derive the receiver operating characteristic (ROC) of the energy detector and show, analytically, the impact of the sliding window length on the energy detection. A learning algorithm is proposed, allowing the Radiobot to independently learn from its past experience in order to optimize its operating parameters. By applying the learning algorithm to the sensing module, we verify, through simulations, the convergence of the proposed algorithm to the optimal solution.