Optimisation de l’utilisation du spectre par les techniques de la radio cognitive dans les systèmes de communication sans fil.

Abstract

The major challenge in the new emerging technologies like IoT and 5G is the obtention of available and enough spectrum resources for their transmissions. Hence, it becomes necessary to optimize the actual spectrum utilization that is relying on the static spectrum allocation, by introducing the dynamic attribution of this scarce resource. CRNs are expected to tackle this issue by enabling the coexistence of secondary users with primary users via heterogeneous wireless architectures and dynamic spectrum access techniques. In this work, two main CR aspects are studied and evaluated. First, spectrum observatory and real database collection were performed to recognize the spectrum behavior and investigate the spectrum availability, for the integration of CR opportunistic networks. These measurements were achieved in cooperation with the ANF- Algeria, between January and February 2020, in two areas, one urban in the North in Constantine, and another rural in the south in Ouargla. The results of these measurement campaigns reveal low resource occupancy, lower than 30.27%, by comparing the occupied instants of each frequency band to its total number of samples, for both areas. In another hand, the impact of the spectrum observatory on the spectrum management strategy preferences is studied. Second, a low complexity spectrum prediction and preallocation system based on an optimized NN model for CR-IoT users is presented. The Bayesian Optimization algorithm was used for the optimization and evaluation of two NN prediction architectures, which are trained on a real spectral occupancy dataset, then compared. Very efficient results have been obtained, high prediction accuracy of 93.5%, a regression coefficient of 0.98, and a reduced MSE of 0.0013. Results show that the considered scheme is efficient in predicting the occupancy rates of different bands within the IoT spectrum resources.