Green cooperative spectrum sensing and scheduling in heterogeneous cognitive radio networks

The motivation behind the cognitive radio networks (CRNs) is rooted in scarcity of the radio spectrum and inefficiency of its management to meet the ever increasing high quality of service demands. Furthermore, information and communication technologies have limited and/or expensive energy resources and contribute significantly to the global carbon footprint. To alleviate these issues, energy efficient and energy harvesting (EEH) CRNs can harvest the required energy from ambient renewable sources while collecting the necessary bandwidth by discovering free spectrum for a minimized energy cost. Therefore, EEH-CRNs have potential to achieve green communications by enabling spectrum and energy self-sustaining networks. In this thesis, green cooperative spectrum sensing (CSS) policies are considered for large scale heterogeneous CRNs which consist of multiple primary channels (PCs) and a large number of secondary users (SUs) with heterogeneous sensing and reporting channel qualities. Firstly, a multi-objective clustering optimization (MOCO) problem is formulated from macro and micro perspectives; Macro perspective partitions SUs into clusters with the objectives: 1) Intra-cluster energy minimization of each cluster, 2) Intra-cluster throughput maximization of each cluster, and 3) Inter-cluster energy and throughput fairness. A multi-objective genetic algorithm, Non-dominated Sorting Genetic Algorithm-II (NSGA-II), is adopted and demonstrated how to solve the MOCO. The micro perspective, on the other hand, works as a sub-procedure on cluster formations given by macro perspective. For the micro perspective, a multihop reporting based CH selection procedure is proposed to find: 1) The best CH which gives the minimum total multi-hop error rate, and 2) the optimal routing paths from SUs to the CHs using Dijkstra\u27s algorithm. Using Poisson-Binomial distribution, a novel and generalized K-out-of-N voting rule is developed for heterogeneous CRNs to allow SUs to have different levels of local detection performance. Then, a convex optimization framework is established to minimize the intra-cluster energy cost subject to collision and spectrum utilization constraints.Likewise, instead of a common fixed sample size test, a weighted sample size test is considered for quantized soft decision fusion to obtain a more EE regime under heterogeneity. Secondly, an energy and spectrum efficient CSS scheduling (CSSS) problem is investigated to minimize the energy cost per achieved data rate subject to collision and spectrum utilization constraints. The total energy cost is calculated as the sum of energy expenditures resulting from sensing, reporting and channel switching operations. Then, a mixed integer non-linear programming problem is formulated to determine: 1) The optimal scheduling subset of a large number of PCs which cannot be sensed at the same time, 2) The SU assignment set for each scheduled PC, and 3) Optimal sensing parameters of SUs on each PC. Thereafter, an equivalent convex framework is developed for specific instances of above combinatorial problem. For the comparison, optimal detection and sensing thresholds are also derived analytically under the homogeneity assumption. Based on these, a prioritized ordering heuristic is developed to order channels under the spectrum, energy and spectrum-energy limited regimes. After that, a scheduling and assignment heuristic is proposed and shown to have a very close performance to the exhaustive optimal solution. Finally, the behavior of the CRN is numerically analyzed under these regimes with respect to different numbers of SUs, PCs and sensing qualities. Lastly, a single channel energy harvesting CSS scheme is considered with SUs experiencing different energy arrival rates, sensing, and reporting qualities. In order to alleviate the half- duplex EH constraint, which precludes from charging and discharging at the same time, and to harvest energy from both renewable sources and ambient radio signals, a full-duplex hybrid energy harvesting (EH) model is developed. After formulating the energy state evolution of half and full duplex systems under stochastic energy arrivals, a convex optimization framework is established to jointly obtain the optimal harvesting ratio, sensing duration and detection threshold of each SU to find an optimal myopic EH policy subject to collision and energy- causality constraints