Abstract

Cognitive ad-hoc networks are able to share primary user frequency bands following certain interference preconditions. For considered cognitive network, cognitive communication is limited by the interference
imposed on the primary user. Probability of channel availability for cognitive nodes for such opportunistic access is determined. Furthermore, this probability of channel availability is used for the performance analysis purpose. A Carrier Sense Multiple Access (CSMA) Media Access Control (MAC) protocol for the cognitive network is considered and for that the embedded Markov model of cognitive nodes is determined. This Markov model is used to determine the average channel access delay, throughput and service rate of cognitive nodes.

This network is further extended to consider multiple frequency bands for cognitive access. For this propose algorithms are proposed to address the channel allocation and fairness issues of multi-band multiuser cognitive ad-hoc networks. Nodes in the network have unequal channel access probability and have no prior information about the offered bandwidth or number of users in the multiple access system. In that, nodes use reinforcement learning
algorithm to predict future channel selection probability from the past experience and reach an equilibrium state. Proof of convergence of this multi party stochastic game is established. Nevertheless, cognitive nodes can reduce the convergence time by exchanging channel selection information and thus further improve the network performance.

To further improve the spectrum utilization, this study is extended to include Multiple-input Multiple-output (MIMO)
techniques. To improve the transmission efficiency of the MIMO system, a cross-layer antenna selection algorithm is proposed. The proposed cross-layer antenna selection and beamforming algorithm works as the data link layer efficiency information is used for antenna selection
purpose to achieve high efficiency at the data link layer.

Having analyzed the cognitive network, to consider more realistic scenario primary users identification method is proposed. An artificial intelligent method has been adopted for this purpose. Numerical results are presented for the algorithm and compare these results with the theoretical ones.