Loading...

Cooperative Spectrum Sensing Based on Learning Techniques for 5G Mobile Communication Networks

Karimpour Fard, Elaheh | 2021

137 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54859 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Behroozi, Hamid
  7. Abstract:
  8. In recent years, with the continuous growth and development of wireless communication systems, the demand for the use of radio spectrum has increased significantly, which has led to problems such as a shortage of spectrum. An effective solution to the spectrum shortage problem is to use cognitive radio. One of the main functions of cognitive radio networks is spectrum sensing and our focus in this study is to examine machine learning and deep learning methods to improve cooperative spectrum sensing. Since channel and noise parameters are not accurate in telecommunication systems, we need to use a method to estimate these parameters. In this regard, the structure of Bayesian with neural network is used to estimate the unknown parameters, and the problem of spectrum sensing is solved. It can be seen that the use of Bayesian structure compared to CNN and LSTM methods between 5% to 15% has helped to better detect the state of the spectrum in a large number of antennas. Next, we upgrade the system and by using the H-learning reinforcement learning method, we increase the efficiency of the spectrum compared to the previously used methods such as Q-learning and SARSA. By making changes to the reward function of this method, we can improve its performance by about 3%, without complicating it. Participatory cooperative spectrum sensing methods are more complex, so they require more energy to detect, and in most cases, energy supply is difficult. Therefore, to increase energy efficiency, the plan of energy harvesting and simultaneous wireless information and power transfer is used. Due to the fact that the information received from the environment is constantly changing, the use of reinforcement learning methods helps to improve the performance of spectrum sensing. Finally, in order to investigate non-orthogonal multiple access systems, we changed the system from orthogonal multiple access systems to non-orthogonal multiple access systems, and we examine the problem of the spectrum sensing and energy efficiency in the form of a non-convex optimization problem and in the framework of rate splitting multiple access using the DDPG deep learning method. The results show that the total power transmitted from the base station in the DDPG method is about 19dBm, while through the DQN and DDQN methods, the transmitted power is 22dBm, which indicates the better performance of the DDPG method.
  9. Keywords:
  10. Cognitive Radio ; Cooperative Spectrum Sensing ; Machine Learning ; Deep Learning ; Energy Harvesting ; Rate Splitting Multiple Access

 Digital Object List

 Bookmark

No TOC