Cooperator Selection Mechanism Based On Reinforcement Learning In Cognitive Radio

With the rapid development of wireless broadband communication and the large applications of Internet of Things, more and more devices need to allocate spectrum resources. The fixed spectrum allocation policy can no longer meet these demands. As an efficient solution to the scarcity of spectrum resources, cognitive radio is received extensive attention nowadays.In cognitive radio system, cooperative spectrum sensing can efficiently improve the detection performance. In cooperative spectrum sensing, there are various types of threats such as jammers, data falsification and denial of service and environment conditions such as sensors in high shadowing zone that lead to unreliable cooperators. Once the fusion center chooses these unreliable cooperators to report their sensing results, it will severely affect the accuracy of the decision. Therefore, how to select reliable cooperators for data fusion is an important issue in cooperative spectrum sensing.This thesis uses TD algorithm in Reinforcement Learning to select the cooperator based on the existing researches. Take minimizing interference to primary user while maximizing the spectrum utilization as the goal function, numbers of report results equal to FC’s decisions as the reword, the FC calculate a suitability parameter for each cooperator and select the proper one according to the suitability list. The influence of different K and M values in "K out of N" rule for miss alarm probability and false alarm probability is also calculated in this thesis. Base on the calculation, we propose to adjust the number of cooperators by these probability requirements. The simulation results show that this algorithm can dynamically adapt to the cooperators’ behavior and with the adjustment of cooperator number, the performance requirement in cognitive radio system can be satisfied.