Medium access control protocol design, analysis and identification in cognitive radio networks.

摘要

Cognitive Radio (CR) is a promising technique to address the spectrum scarcity issue by enabling the unlicensed network users (SUs) to dynamic access the spectrum holes. This thesis designs Medium Access Control (MAC) protocols for SUs to utilize the unused spectrum without interfering the licensed network users' (PUs) transmissions; analyzes the MAC performance based on the queuing theories; and identifies the PU's and SU's MAC protocol types using machine learning techniques in order to implement smart cognitive radio. In the MAC protocol design part, the unused time slots in primary TDMA networks are considered as spectrum holes for CR transmissions. A Cognitive Carrier Sensing Multiple Access (Cog-CSMA) protocol and a Cognitive Packet Reservation Multiple Access (Cog-PRMA) protocol are proposed for Bernoulli random secondary packet arrivals and bursty secondary packet arrivals scenarios respectively. The proposed Cognitive MAC protocols achieve desirable throughput and delay performance in CR networks, meanwhile avoid the potential inference to the primary TDMA users. In the second part of this thesis, the Tagged User Analysis (TUA) is investigated for cognitive MAC protocol performance analysis, which utilizing the classic queuing theories to simplify the computational complexity. One MAC protocol, cognitive slotted ALOHA, is used as an example to verify the TUA. Compared with the traditional Markov chain analysis, the proposed TUA achieves similar analytical results but significantly reduces the computational complexity. In the identification part of this thesis, MAC protocol identification based on machine learning is investigated. Four MAC protocols, ALOHA, slotted ALOHA, TDMA and CSMA are considered in the identification process. Moreover, a feature extraction scheme is proposed in this thesis. The power and channel states features are extracted from the received signal in order to perform MAC protocol identification, which in turn facilitates the design of cognitive MAC protocols to use the spectrum holes in a more efficient way.