|Ph.D Student||Oklander Boris|
|Subject||Modeling Cognitive Radio Systems|
|Department||Department of Electrical Engineering||Supervisor||Professor Moshe Sidi|
|Full Thesis text|
In order to provide stable communication sessions over dynamic communication environment, cognitive radio (CR) relies on its cognitive capabilities. These capabilities allow CR to allocate its resources dynamically and to make decisions that would optimize its performance. In this work we propose a stochastic framework for modeling the adaptive behavior of CR. Our model includes the interdependent cognitive and communication processes and their mutual impact on the CR’s performance. The communication processes include network dynamics, packet arrivals, queueing and transmissions. The cognitive processes consist of sensing and estimation of the environments state, planning and dynamic resource allocation. The proposed framework is analyzed using tools from queueing theory and CR’s performance is evaluated. Then, model-based analysis is used to optimize CR’s decision-making process. We design an efficient strategy for accessing the vacant spectrum bands and managing the transmission-sampling trade-off. In order to cope with the high complexity of this problem the policy search uses the stochastic optimization method of cross-entropy. The optimized performance measures are better by orders of magnitude over performance obtained by a greedy-policy. Additionally, we examine the sensitivity of CR’s performance to different types of errors. Finally, a prediction-based spectrum access method is proposed. The proposed method is embedded into CR’s analytical model. We show that this approach has a great impact on the average waiting time as well as on system stability.