Multi-dimensional Anderson-Darling statistic based goodness-of-fit test for spectrum sensing

In this paper, we propose a multi-dimensional extension of the Anderson-Darling statistic based goodness-of-fit lest for spectrum sensing in a cognitive radio network with multiple nodes. A technique lo evaluate the optimal detection threshold that satisfies a constraint on the false-alarm probability is discussed. Assuming stationary and known noise statistics, we show that this detector, called as the K-sample Anderson-Darling statistic based detector, outperforms the well-known energy detector under various practically relevant primary signal models and channel fading models, through extensive Monte Carlo simulations.