الفهرس 
NTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 71 |  |  |
| | | from | 71 | | |
Abstract
Cognitive Radio Networks are getting higher momentum. With lots of routing protocolsthat exist in the literature, to the best of our knowledge, none exploits a more generalcognitive processing, instead of basing the decision only on the local view of the locallysensed channels. By the general cognitive processing it is meant: (A) applying thecomplete intelligence cycle (observe, learn, act) on end-to-end network variables, and(B) relating those network variables to network-wide goals.The proli_c usage of wireless spectrum by pervasive wireless communication technolo-gies. The current static management of frequency spectra, which results in an unevenusage patterns. The emergence of the CR (Cognitive Radio) enabling technologies (SDR(Software De_ned Radio), USRPs (Universal Software Radio Peripheral), ..) which inturn is a _eld of study that is beyond the scope of this one, are all examples of motivat-ing research in Cognitive Radio Networks. Current wireless networks are regulated bygovernmental agencies according to a _xed spectrum assignment policy. Licensees aregranted the rights for the use of various, often relatively small, frequency bands on a longterm basis over vast geographical regions. In recent years, the huge success of wirelessapplications has caused an exponential increase in requests to regulatory authorities forspectrum allocation. This has stimulated the reassignment of bands used by old legacysystems, or by technologies that have transitioned to digital communications (e.g. Dig-ital Dividend, which is due to switching over from analogue television signals to digital.This switchover will free up a signi_cant amount of spectrum (primarily in the UHFband)).However, even with these reassignments, the spectrum shortage persists. Part of theproblem is that spectrum usage is not uniform. On the one hand, the use of wireless tech-nologies operating in unlicensed bands, especially in the ISM band, has been proli_c witha wide range of applications developed in di_erent _elds (e.g. WLANs (Wireless LocalArea Networks), mesh networks, PANs (Personal Area Networks), body area networks,WSNs (Wireless Sensor Networks), etc.), which caused over-crowding in this band. Onthe other hand, the usage of licensed spectrum is quite uneven and depends heavily onthe speci_c wireless technologies, their market penetration, and the commercial successof the operators to which the frequencies have been assigned.A novel routing solution is proposed. It generalizes the cognitive process (perceiveing, learning, adaption) from the local Cognitive Radio to a more general CognitiveNetwork, this is achieved by designing routing agents to be autonomous routing agentsthat perform the cognitive cycle based on Q-based reinforcement learning technique tominimize the overhead of information transfer. As compared to the most relevant work,the proposed protocol can provide better quality of service (700% better response timeand 67% better delay) under more loaded network conditions (20 in-parallel ows), andis scalable in di_erent network scenarios.