Optimization for Prediction-Driven Cooperative Spectrum Sensing in Cognitive Radio Networks

Empirical studies have observed that the spec-trum usage in practice follows regular patterns. Machine learning (ML)-based spectrum prediction techniques can thus be used jointly with cooperative sensing in cognitive radio networks (CRNs). In this paper, we propose a novel cluster-based sensing-after-prediction scheme and aim to reduce the total energy consumption of a CRN. An integer programming problem is formulated that minimizes the cluster size and optimizes the decision threshold, while guaranteeing the system accuracy requirement. To solve this challenging optimization problem, the relaxation technique is used which transforms the optimization problem into a tractable problem. The solution to the relaxed problem serves as a foundation for the solution to the original integer programming. Finally, a low-complexity search algorithm is proposed which achieves the global optimum, as it obtains the same performance with exhaustive search. Simulation results demonstrate that the total energy consumption of CRN is greatly reduced by applying our clustered sensing-after-prediction scheme.