Implementation and Benchmarking of CUSUM Algorithm for Spectrum Sensing under Low SNR Regimes on ANRC Cognitive Radio Testbed

Most of the research in dynamic spectrum access and cognitive radio network is directed towards the formulation of methods for spectrum access, management etc., that too, mostly in form of theoretical models with some simulated results to back their work. Many algorithms though novel in approach are designed with certain assumptions which are sometimes very hard to realize in practical scenarios. Therefore implementation of certain class of algorithms on platforms with real world environments is even more challenging, which perhaps is very important to test the hypothesis on which the results are presented by the designer/s of these algorithms. In this paper we discuss on the practical implementation of a sequential change detection algorithm - CUSUM (Cumulative Summation). We have implemented and characterized one of the several versions of CUSUM for single node sensor as the first step for benchmarking and also to unearth the complexities involved in its implementation on a hardware platform. We have used the ANRC´s hybrid Cognitive Radio testbed for the implementation, measurement and characterization of the same. We have found that the sequential change detector based algorithm is more robust detector than Snapshot ED under very low SNR regimes and through measurement and analysis we conclude that the online version of the same can avoid the SNR wall problem altogether.