Optimal dynamic sensing policy for cognitive radio network: A POMDP model based method

In the cognitive radio(CR) network, the secondary user(SU) are allowed to utilize the frequency channels of the licensees primary user(PU) when these channels are not currently being used. To support spectrum reuse functionality, the SU are required to sense the channel to detect the PU activity accurately for each time slot. There are two parameters associated with spectrum sensing: probability of detection and probability of false alarm. Given the probability of detection, the long sensing time reduce the false alarm probability, however the left time used for transmission is short. There is a trade off between the sensing cost and successful transmission reward. When we combined the PU activity property with the discounted utility of the SU, how to improve the SU utility under the unreliable spectrum sensing situation is still a challenge problem and rarely be studied. In this paper we propose a new optimal dynamic sensing policy, called sleep-sensing-transmission strategy based on partially observable Markov Decision Process(POMDP). At the beginning of each slot, the SU can choose three different possible actions: Sleep, spectrum sensing and then access the channel based on the sensing result or directly transmission to maximize the expected discounted utility in the long run. In this model we have shown that the optimal strategy has a simple threshold structure, and the simulation result validate the efficiency of our policy compared with the traditional spectrum sensing and access mechanism.