Performance Evaluation of Machine Learning-Based Techniques for Spectrum Sensing in Mobile Cognitive Radio Networks

Abstract

Communication technologies evolve drastically in recent years. These recent developments in communication technologies inevitably lead to increasing demand for communication technologies. They also lead to allocating progressively communication channels between transmitting and receiving elements. However, the scarcity of spectrum began to appear with the accelerating rise in the usage of various communication technologies and the preservation of traditional channel access methods. That is what is called the spectrum under-utilization and frequencies crisis. Cognitive Radio (CR) is an innovative solution developed to avoid these obstacles mentioned above. It deals with the surrounding environment and determines the appropriate communication parameters. The CR life cycle goes through many tasks. Spectrum sensing is a key task in this cycle that gains significance where the spectrum holes can be detected during this task. Spectrum sensing task is needed in both non-cooperative spectrum sensing (Non-CSS) or cooperative spectrum sensing (CSS) modes, whereby the secondary users (SUs) cooperate to determine channel state. The deployment of machine learning-based spectrum sensing techniques in CR networks has been attracting researchers in recent years. The idea is to provide the network with some intelligence to enhance the spectrum sensing process and thus the possibility of accurately detecting PU activity. Examples of these techniques are the energy detection-based, the covariance matrix-based, and Machine learningbased techniques. In this thesis, firstly we study and compare the performance of the Non-CSS, the And-based, the Or-based techniques, as well as the KMeans-based ML technique in stationary CRNs. In contrast to the majority of published research, we examine the performance of that mentioned spectrum sensing techniques in mobile CRNs. Also, we try to grasp the effect of the fading channels on the sensing performance. Moreover, we try to find the optimal parameters that can improve the performance of various spectrum sensing techniques. Finally, we investigate the circumstances in iii which KMeans-based spectrum sensing techniques introduce superior to traditional techniques. Stationary and mobile CRNs were simulating using the third version of the network simulator (ns3 simulation). The result showed the effect of the communication channel on the spectrum sensing performance. Difficult conditions such as noise and fading effects on the transmitted signals lead to a notable decrease in sensing performance. Also, the results generally revealed that spectrum sensing techniques provide better performance in stationary networks. However, the results showed we need above three SUs and about 1500 samples to reach an acceptable performance level in mobile CRN. In addition, the results showed that the KMeans-based technique slightly outperforms the Or-based technique, especially in highly-noisy environments and under severe fading channels.