Tests using real radio equipments were also performedagainst pseudo random and reactive jammers. The measured results confirm the same conclusions as simulationresults; The channel selection based on the proposedcooperative algorithm outperforms learning without cooperation which also outperforms the best channel selectionwithout learning. However, the real USRP measurements are not equal to MATLAB simulation values. This is due tothe implemented time division multiplexing scheme thatneeds a processing time for the blind reception of thepackets or the acknowledgments. In MATLAB, the simulation time is different from the real time and neither theCR nodes nor the jammers need a processing time aspresented in Fig. 2.