الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 209 |  |  |
| | | from | 209 | | |
Abstract
The tools of stochastic modeling and stochastic optimization are powerful tools for representing and optimizing different systems, especially in random environments. In such environments, some system components could behave randomly, e.g. amount of resources available, partial observability of part of the system state, etc. Markov decision process (MDP) and its variations, e.g., partially observable MDP (POMDP) and mixed observable MDP (MOMDP) provide a solution for modeling and optimization in such environments. The concepts of dynamic programming are used to solve the MDP problems. However, partial observability and high dimensionality of the system state introduce high computational complexity to the solution. High computational complexity is considered a major challenge, especially when the algorithms are applied to small devices with limited computational power. These stochastic tools are applied to cognitive radio networks (CRNs) with energy harvesting. In these CRNs, spectrum, energy, and secrecy efficiency are to be optimized in a random environment. For spectrum efficiency, CRNs target the problem of the scarce spectrum bands by increasing the channel utilization through dynamic spectrum access. For example, the unlicensed secondary users (SUs) are allowed to use the licensed bands opportunistically without affecting the licensed primary users (PUs). Such a wireless network can be found in many daily applications such as body measurement devices, medical equipment, wearable devices, wireless sensors networks, etc. Energy efficiency could be enhanced by harvesting energy from ambient sources (e.g., wind, solar, electromagnetic, etc.). This is especially useful for mobile devices or for devices installed in hard-to-reach environment. On the other hand, the continuous connectivity between these devices is considered as one of the main targets for the next generation networks. As a result, the security issue is of great interest in such networks. The unpredictable availability of both spectrum and energy sources is the main challenge in optimizing the decision by the users of such networks. In this work, an improved spectrum sharing framework is proposed to enhance wireless network spectrum utilization for battery-limited energy harvesting devices. This framework takes into account the random environment behavior such as the dynamic spectrum/energy availability, wireless channel conditions, etc. Therefore, we propose to deploy stochastic optimization techniques such as MDP and its variations to select the optimum action to be taken by the network user. Moreover, due to the limited capacity of the devices’ batteries used in such networks, energy harvesting technology is merged in the proposed framework as a promising solution to increase the operating lifetime of the devices. On the other hand, the huge development of the computational capabilities raises questions about the efficiency of the traditional authentication methods which are based on mathematical foundations. Therefore, we enhance the security of the proposed framework by using the recent physical layer security techniques against eavesdropping. The aforementioned approaches used to tackle the challenges are expected to enhance spectrum utilization, energy efficiency, and secrecy capacity. Moreover, efficient solution techniques are used to ease the complexity of the MDP solution due to partial observability of some system states and the high dimensionality of the problem. Based on a literature review, a basic research gap is identified and studied. The basic gap is to optimize the SUs performance in a CRN. In this CRN, the SUs access the network in hybrid interweave/underlay mode, and they have the ability to harvest energy from multiple sources. First, throughput of those SUs is maximized and the infinite-horizon optimal policy is constructed using MOMDP considering the random environment. Next, the security of the basic model is investigated and enhanced using artificial noise. Finally, the basic model is upgraded to consider more practical issues, e.g., the different quality of service requirements of the SUs and the channel fading model.