Lessons Learned from Long-Term and Imperfect Sensing in 2.4 GHz Unlicensed Band

Abstract

Accuracy of spectrum sensing affects the decision making operation of cognitive radio. In order to achieve meaningful results, in related experimental and simulation work, realistic wireless environment representation is a necessity. Existing spectrum occupancy models range from simple additive white Gaussian noise to elaborate, based on large scale wireless spectrum measurements, but universal models are not available. Creating such a model for unlicensed bands would be particularly difficult, if not impossible, because of its unpredictability and inherent dynamics. On the other hand, our experience shows that using real-life, relatively low-resolution, data collected using inexpensive spectrum analyzer provides insight consistent with observations made with more sophisticated setups, preserves more nuances than simple models, and could be a viable alternative to spectrum occupancy modeling.