The Wideband Spectrum Detection Based On Adaptive Q-learning

Cognitive radio is a new technology which is able to automatically sense theenvironment and detect the spectrum holes. Due to the actual environment isdynamic, so the adaptive dynamic environmental changes of spectrum detection isone of the difficult points of the present study. This article revolves around how toadaptive fast and accurate detection spectrum holes.Due to long sensing time, excess energy consuming and incapability of adaptive,this paper proposes a algorithm of spectrum sensing based on clustering cooperativeQ-learning in large-scale cognitive radios systems.The algorithm uses clustermechanism, divided the large-scale environment into small ones,using thecollaborative q-learning in the divided clusters.Obtain the optimal threshold ofspectrum detection by continuous processes of “trial and error” in the interactionbetween agents and environment.By this way,the system will have the ability ofautonomous learning.The experimental results show that: the detection performanceof system improved significantly.In traditional compressive sampling approach, it’s always assumed the sparsityof a signal is known. However, we always can’t get this in Cognitive Radio (CR)network. This makes a great barrier to the practical usage of compressive sampling.This paper develops an Q-learning Adaptive Compressive Sampling (QACS)approach for wideband signals. It doesn’t need the sparsity as a priori knowledge.Also, once the sparsity changes, it can adaptively change the proper sampling ratethrough comparing the different results recovered by different sampling rates. Whenthere is no difference between the results recovered by two different sampling rates,the smaller sampling rate of these two is the proper sampling rate to recover theoriginal information of the signal exactly. The experimental results show that: thedetection performance of system improved significantly.