Optimization of NoutofK Rule
Banavathu, N R and Khan, Mohammed Zafar Ali (2018) Optimization of NoutofK Rule. PhD thesis, Indian institute of technology Hyderabad.
Text
Thesis_Phd_EE_4614.pdf  Submitted Version Restricted to Repository staff only until December 2020. Download (4MB)  Request a copy 
Abstract
The explosive growth in wireless systems and services has led to spectrum scarcity in wireless communications due to everincreasing demand for higher data rates. The shortage of useful radio spectrum is due to static allocation to specific services and rigid regulation of the spectrum usage rather than efficient utilization of the radio spectrum. Cognitive radio (CR) technology has been proposed to alleviate the spectrum shortage problem and the spectrum underutilization of current radio spectrum by allowing the cognitive users (CUs) to access spectrum of the licensed or primary user (PU) under sufficient protection to the PU. Therefore, spectrum sensing is a fundamental component for the CU to accurately detect the activities of the PU. However, spectrum sensing using single CU results in poor detection performance due to mulipath and hidden terminal problem. Therefore, cooperative spectrum sensing (CSS) is proposed to enhance the detection accuracy by taking the advantage of spatial diversity in the multiple CU observations. In this thesis, we consider decision based CSS, where in each CU makes a binary decision on the activity of the PU and the decisions are reported to the FC, where they are combined using NoutofK rule, i.e., at least N CUs must favour for the presence of the PU out of K CUs. Most existing works in the literature on NoutofK rule are for homogeneous CR networks, where CUs are assumed to have identical local false alarm and identical detection probability. However, in practice, the sensing accuracy will vary since the CUs are at different distances from the PU. In this thesis, we consider heterogeneous CSS using the NoutofK rule, wherein spatially dispersed CUs operate at different local false alarm, detection probabilities and detect the activities of the PU using the NoutofK rule. We obtain the generalized expressions for the global false alarm and global detection probabilities for the NoutofK rule for the heterogeneous CR network in the presence of control channel errors and the proposed solution specializes to various existing results. We then propose NeymanPearson (NP) test which maximizes the global detection probability of the NoutofK rule for a given target global false alarm probability at the FC. We show limitations on the target global false alarm probability due to control channel errors. We then consider the optimization of NoutofK rule for the heterogeneous CR network under Bayesian test and show that several existing works are special cases of the proposed solution. We also obtain a most generalized expression for optimal K for the homogeneous CR network and show that various results are special cases of the proposed solution. A joint optimization problem is formulated to optimize both N and K under Bayesian test. The performance of the CSS obtained using joint optimized values of N and K results in significant improvement. Finally, we propose, optimization of NoutofK rule for maximizing the average channel throughput (ACT) of the heterogeneous CR network. We present asymptotic expressions for the optimal N and the optimal K by maximizing the ACT under adequate protection to the PU. Note that the mathematical results derived in this thesis are general and are applicable to any detector used in CSS. However, we choose the energy detector (ED) as an example to verify the theoretical findings
[error in script]IITH Creators: 



Item Type:  Thesis (PhD)  
Uncontrolled Keywords:  Discrete Optimazation, NeymanPearson Test, Baysein Test, heterogeneous Network  
Subjects:  Electrical Engineering  
Divisions:  Department of Electrical Engineering  
Depositing User:  Team Library  
Date Deposited:  05 Dec 2018 06:19  
Last Modified:  05 Dec 2018 06:19  
URI:  http://raiith.iith.ac.in/id/eprint/4614  
Publisher URL:  
Related URLs: 
Actions (login required)
View Item 
Statistics for this ePrint Item 