Utilizing Q-Learning to allow a radar to choose its transmit frequency, adapting to its environment

Recent research show that utilization of knowledge of the environment can allow a radar system to adapt its processing to improve its performance. Furthermore, a radar system that utilize both a-priori and measured knowledge in an adaptive close loop manner could seem to be cognitive of its environment, able to adapt to changes to optimize performance. Reinforced learning could play a vital role as part of such a closed-loop cognitive radar system. The Q-Learning algorithm is hypothesized to be useful for this cognitive radar domain. This paper investigates the problem of adaptively choosing the radar transmit frequency through application of Q-Learning on measured radar data. A comparison is made against other frequency selection algorithms and its shown that Q-Learning manages to learn a good strategy to adaptively select radar transmit frequency, mostly outperforming the other methods tested in the scenario investigated here.