Learning multi-channel power allocation against smart jammer in cognitive radio networks

We model the power allocation interaction between a cognitive radio and a jammer as a two-player zero-sum game. First, we determine the power allocation strategy for the cognitive radio using a modified version of the Q-learning algorithm against fixed jamming strategies. The learned anti-jamming strategy will be compared to the common waterfilling technique. Then, we consider the power allocation game using Q-learning for both the cognitive radio and the jammer. The learned strategies will be compared to the Nash equilibrium found under the assumption of perfect knowledge. Finally, we consider the real scenario of a jammer with imperfect information.