Sequential learning for passive monitoring of multi-channel wireless networks

With the requirement for increasing efficiency of wireless spectrum usage, the cognitive radio technique has been emerging as an important solution. Passive monitoring over wireless channels in cognitive radio is an innovative approach in which the system attempts to locate channels with the highest activity over time. A huge amount of work has been contributed to this field when the reward of each channel is identical to observers. However, when the reward is different over observers, these algorithms perform poorly. In this thesis, we challenge this problem by considering this correlation as part of the reward. We develop one optimal online learning algorithm when a switching cost exists in the system. We also propose three approximation algorithms with competitive computation complexity but still guarantee to obtain a constant amount of reward compared to the optimal case. Theoretical analysis and simulation are conducted to prove the effectiveness of these approaches.