Spectrum Sensing Using Hybrid Gravitational Search Genetic Algorithm Based Predictor Model in Cognitive Radio Networks
17 Pages Posted: 7 Mar 2018
Date Written: November 15, 2017
The exponential growth of wireless users and devices increases the demand for frequency spectrum. The radio spectrum is considered as a precious natural resource that is available in a limited range for the use of wireless services. Hence the primary challenge is to design efficient spectrum allotment policies that utilize the spectrum in a better way. The frequency spectrum has been regulated and assigned to the service provider by ITU. Only 70% of the spectrum bands are utilized efficiently while 30% are underutilized by the licensed users. These spectrum holes can be used by the users in an opportunistic manner through Dynamic Spectrum Access (DSA) technique. Spectrum holes of the licensed users have to be utilized by an unlicensed user without causing interference to the licensed user. Among the various functionalities of cognitive radio, spectrum sensing plays a significant role in determining the availability of spectrum holes. The Time and energy consumed by the spectrum sensing phase can be reduced by incorporating prediction before sensing the entire spectrum range. The prediction helps in predicting holes in near future based on the current and the past history. By employing an accurate and reliable prediction model, the secondary user’s energy can be conserved, so that only idle channels which are predicted by the model be sensed by the sensing module. The proposed work presents a hybrid predictor scheme that uses Gravitational Search Genetic Algorithm (GSGA) in Backpropagation Neural Network to forecast the Primary User (PU) channel occupancy status. The proposed approach is simulated using MATLAB and the results show high accuracy when compared with Neural Network model based on Backpropagation.
Keywords: Dynamic Spectrum Access, Spectrum Selection, Prediction model, Hybrid Gravitational Search algorithm, Genetic Algorithm
Suggested Citation: Suggested Citation