الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 114 |  |  |
| | | from | 114 | | |
Abstract
Cognitive radio is an escalating technology with the aim of enhancing the utilization of theavailable spectrum. The unlicensed secondary users (SUs) are allowed to transmit their dataon the licensed bands of the authorized primary users (PUs). Challenges arise due to thetemporary use of the spectrum by the SUs as they have to vacate the channel upon the arrivalof the PU and select a target channel to continue their unfinished transmission in a processcalled spectrum handoff.This study is concerned with the proactive target channel selection approaches, namely, thefixed sequence and the probabilistic target channel selection strategies where the target channelis predetermined before starting the transmission based on the long term statistics. TheIEEE 802.22 wireless regional area networks (WRAN) standard suggests two proactive targetchannel selection approaches, namely, ”always stay” and ”always change”. The formerstates that the SU stays at his current channel every time he gets interrupted by the PU. However,the latter states that the SU has to change his current channel after each interruption.These strategies are studied and proven of efficiency in many network situations. Moreover,the pre-emptive resume priority (PRP) M/M/1 queuing model is used as a channel modelfor its ability to meet a lot of handoff requirements in cognitive radio networks. Networkbalancing, delay, and optimum target channel selection approach are investigated. Moreover,the cognitive radio network is studied for different loading conditions (homogeneous andnon-homogeneous) and for different channel conditions (identical and non-identical).Firstly, the analytical formulae for the extended data delivery time for the SUs are derived inthe case of identical channels. The derived formulae can evaluate the extended data deliverytime for any fixed sequence or probabilistic transition matrices. The analytical formulae obtainedfor the extended data delivery time and also for the channels’ busy probabilities areused to evaluate and compare the latency performance and load-balancing of the connectionbasedspectrum handoff techniques in cognitive radio networks. Moreover, a new quantitativemeasure for load balancing is introduced. The results reveal the advantage of theprobabilistic approach (based on the total arrival rate to each channel) in load balancing interms of the variance of the channels’ busyprobabilities.Secondly, a generalized model is proposed for the case of non-identical channels’ network.The analytical formulae for both the fixed sequence and the probabilistic target channel selectionapproaches are derived. The new generalized model is tested under many networkconditions of homogeneous / heterogeneous loads and identical / non-identical channels. Thenumerical results are found to match the simulation results, which verifies our generalizedmodel. Moreover, the formulae are found to be reducible analytically to the special case of
homogeneous load and identical channels network.Finally, an optimal target channel selection approach for the proposed non-identical channels’cognitive radio network is introduced. Both the fixed sequence and the probabilisticapproaches are optimized using meta-heuristic methods of genetic algorithm and particleswarm, respectively, in the seek of finding an optimum approach to minimize the extendeddata delivery time of the SUs. Based on the optimization results for different networkcases, the behaviour of the optimized transition matrices fixed sequenceandprobabilisticapproach) motivated us to propose a suboptimal fast method to determine a near optimalchoice for the target channel.