This thesis presents novel architectures and algorithms to improve the spectral efficiency of Cognitive Radio (CR) systems using 3-Dimensional Spectrum Sensing (3DSS) and Support Vector Machines (SVMs). Spectrum sensing requires CR nodes to co-operate with each other to sense the presence of a Primary User (PU) to avoid the hidden node problem (Ian F. Akyildiz, 2009). This thesis presents a novel scheme for assigning the resources for sensing and communication based on the distance estimated using path loss curve. The thesis proposes controlled co-operative spectrum sensing to enhance the spectral efficiency and throughput of the CR network. In the proposed controlled co-operative spectrum sensing, time, frequency and distance of CR nodes to fusion centre are assigned through a fusion centre to enhance the effective spectrum sensing of the CR system. Unlike the conventional CR systems where the spectrum sensing is in frequency domain only, this thesis proposes 3DSS in space, time and frequency domains to reduce the overall probabilities of false alarm and missed detection. Further, MANET is overlaid on a heterogeneous network to offload signalling and functions of the fusion centre. Overlaying MANET on the infrastructure based heterogeneous network makes link establishment and maintenance more robust and hence improves the success of CR deployment on the MANET. This thesis proposes 3DSS routing algorithm in CR enabled MANET (CRMANET) to improve the throughput of CRMANET network. The 3DSS and the 3DSS routing algorithm are used to update route tables used for link routing in CRMANET. Stochastic Gradient Descent (SGD) is used to perform error correction to the route table of CRMANET. This thesis provides a framework for the adaptive update of the route table using SGD. SGD provides error correction through error backpropagation mechanism over route table of MANET and hence it is far more efficient than static updates based on the long-term feedback of the network. Since characteristics of wireless channel change rapidly according to location, time and frequency usage, a 3-dimensional SGD implemented in time, frequency and space domains provides better routing error correction. Simulation results reveal 200% increase in spectral efficiency over the existing methods of spectrum sensing deployed in CRMANET.
|Date of Award||Jun 2020|
|Supervisor||Vasile Palade (Supervisor)|