Anti-jamming transmissions with learning in heterogenous cognitive radio networks

This paper investigates the interactions between a secondary user (SU) with frequency hopping and a jammer with spectrum sensing in heterogenous cognitive radio networks. The power control interactions are formulated as a multi-stage anti-jamming game, in which the SU and jammer repeatedly choose their power allocation strategies over multiple channels simultaneously without interfering with primary users. We propose a power allocation strategy for the SU to achieve the optimal transmission power and channel with unaware parameters such as the channel gain of the opponent based on reinforcement learning algorithms including Q-learning for and WoLF-Q. Simulation results show that the proposed power allocation strategy can efficiently improve the SU's performance against both sweeping jammers and smart jammers with learning in heterogenous cognitive radio networks.