Deep Q-Learning Based Beamforming in the Presence of a Jammer

Upcoming protocols and next generation mmWave communication standards require effective beamforming capabilities to compensate for the path loss suffered by high frequencies carriers. In this paper we explore an alternative to conventional beamforming in the form of reinforcement learning based methods. Instead of using using a lookup table based approach we investigate a reward maximization policy and formulate the Q-learning and the deep Q-learning framework framework to carry out beamforming steering while avoiding a jammer signal.