Research On Multi-channel Dynamic Spectrum Access Algorithms In Cognitive Radio Systems

Today’s expanding demand for wireless services has necessitated cognitive radiotechnology which overcomesthe limitations of the conventional static spectrumallocation policy. As a giant breakthrough in enabling a more efficient use of limitedspectrum resources, dynamic spectrum accessing has been one of the key researches inwireless communication.Spectrum sensing at the physical (PHY) layer to detect theactivitiesof licensedusers (or primary users) is one of the main challenges of cognitive radio. Secondaryusers always have limited spectrum sensing capability at the PHY layer in the sense,they are unable to distinguish between the activities (i.e., spectrum access) of theprimary user and a secondary user. It will cost greatly to setup central controllers or byusing dedicated control channels in existing work. Thus we apply the reinforcementlearning (RL) method to DSA of CRN, and extend the MAC protocol with memoryfrom the condition of single channel to multichannel in CRN, then propose twomultichannel algorithms that based on RL under the condition of totally distributed.Firstly, we give concepts and background of cognitive radio and dynamic spectrumaccessing, and introduce several classic dynamic spectrum access strategies and MACprotocols for a cognitive radio network.Secondly, we extend the MAC protocol with memory from the condition of singlechannel to multichannel in CRN. Through mapping the Q-learning to the DSA in CRN,we propose Q learning based multichannel DSA algorithm with memory. Simulationresults show that the algorithm can achieve high channel utilization under the conditionof limited spectrum sensing.Thirdly,according to a randomization based mixed strategy, this paper proposesaexploration and exploitation trade-off multichannel DSA algorithm with memory byanalysing the trade-off between selecting the best channel or considering the collisionamong cognitive users in the learning process. Simulation indicates that the algorithmcan eventually overcome the limitation of cognitive users’ spectrum sensing capability at the PHY, and settle down to the M-best orthogonal channels with high probability.Finally, we use OPNET Modeler to build a dynamic spectrum sharing (DSS)simulation platform to realize the computer simulation model of dynamic spectrumaccess strategy. The platform is scalabletoevaluate the performance ofdynamic spectrumaccessalgorithms anddynamicspectrum sharingMACprotocols.