Study On Cognitive Radio Core Technologies Toward The Guarantee Of Quality Of Service

Radio spectrum (hereinafter referred to as spectrum) is a type of scarce andprecious natural resource in wireless communications, and most of the exploitedspectrum have already been assigned. With the fast development of wirelesscommunications services, there has been a rapid growth of demand on spectrum inrecent years, which causes the problem of spectrum shortage to be acute. So, spectrumhas been a new bottleneck of the development of wireless communications. Recently,cognitive radio (CR) technology has become an important research hot topic in the fieldof wireless communications since it can effectively deal with the current problem ofwireless spectrum scarcity. The design and implementation of CR technology are facinga lot of challenges although its basic idea seems to be very simple. To promote itspractical realization, there has been much research works to be carried out, especially inthe respect of QoS (quality of service) guarantee.This paper is mainly focused on the study of some parts of core technologiestoward the guarantee of QoS, namely the framework model toward the guarantee ofQoS, spectrum forecast, spectrum access, channel allocation, and packet scheduling.The main works of this paper is as follows.(1) In the respect of the framework model toward the guarantee of QoS, the keyfactors influencing QoS in CR environments are analyzed, the strategies forguaranteeing QoS are explored, and a model for the use of dynamic spectrum isdesigned in term of environment requirements and flexibility, in which some existingmature technologies can be also used. The following works of this paper are based onthis model.(2) As for spectrum forecast, based on the analysis of spectrum historic information,the spectrum forecast performance of the multi-variable-grey-model (MGM) and theadaptive MGM (AMGM) are evaluated, and two improved AMGM named eeAMGMand eeeAMGM for spectrum availability forecast are presented. The simulation resultsshow that eeAMGM and eeeAMGM outperform AMGM in the performance ofspectrum availability forecast, such as error degree, accuracy, reliability, etc. In addition,a mechanism for forecasting the remained length of the duration of ON state and OFFstate is designed by introducing the theories of chaos and neural networks, and thesimulation results validate its effectiveness. (3) As far as spectrum access is concerned, a simulation model of the process ofdynamic spectrum access in CR networks is built by the use of queuing theory withcompulsory priority, and the theoretic analysis with Markov theory is given. Accordingto the analysis, the most favorable conditions for the access of secondary users areobtained. Based on the access conditions, an admission control approach is designed,which can adaptively admit or refuse secondary uses in terms of the actual conditions ofspectrum holes and QoS requirements.(4) So far as channel allocation is concerned, a novel channel allocation model forminimizing allocation risk is proposed, and then a channel allocation approach namedMHR for minimizing handoff risk is designed. MHR overall plans multiple channelallocation requirements and also considers multiple data channels together, includingboth idle and busy channels. In addition, two improved MHR named MHR-TCS-NPwithout preempt and MHR-TCS-PM with preempt for supporting two classes of serviceare designed by redefining handoff risk.(5) In the case of packet scheduling in CR networks, the theoretic analysis ofM-LWDF fairness is given, which shows that M-LWDF fairness is relevant to channelcondition, packet’s arrival process and the ratio of QoS requirements of different servicequeues. Given QoS requirements and other parameters related to channel model andpacket’s arrival process, it is merely relevant to the ratio of the number of users in theservice queues. Based on the theoretic analysis of M-LWDF fairness, an enhancedM-LWDF algorithm named EM-LWDF is proposed and demonstrated. EM-LWDF isstrictly designed in light of QoS requirement fairness criteria. Its fairness is almost notrelevant to the ratio of the number of users in the service queues and the theoreticalvalue of fairness index is equal to1.