Prediction-Based Throughput Optimization for Dynamic Spectrum Access

Cognitive radio (CR) for dynamic spectrum sensing and access has been a hot research topic in recent years. To avoid collision with the primary users, secondary users need to sense the channels before transmitting on them, which is referred to as sensing time overhead. Our previous work shows that the spectral correlations between the channels within the same service are sufficiently high for accurate prediction, which can further be used to reduce the sensing time. With such motivation, in this paper, we propose a new definition, i.e., channel availability vector (CAV), to characterize the state information of a group of licensed channels by introducing spectrum prediction while focusing on the scenario of a single secondary user with multiple channels and leverage it by formulating the throughput optimization problem as a Markov decision process, which is further solved by our optimal sensing scheme and verified with the real spectrum measurement data. The results show that our prediction-based sensing scheme outperforms one existing work.