Multichannel Power Allocation Game against Jammer with Changing Strategy

The antagonistic game between a cognitive transmitter and a smart jammer is considered. The cognitive transmitter divides its transmitting power between several concurrent channels trying to maximize their total capacity. The goal of the game is to adapt the cognitive transmission strategy to the changing strategy of the smart jammer. We propose the modification of the classical Q- learning algorithm by adding memory component for determining the strategy of the cognitive transmitter power allocation among several channels under different scenarios of the jammer behavior. The proposed algorithm successfully handles different jammer strategies and adapts to their changes.