Frequency hopping methods for radar anti-jamming

Sammanfattning

Three approaches for Doppler radar are compared with respect to anti-jamming: A non-adaptive one which maintains a constant transmission frequency, an simple adaptive one that changes frequency based on current interference on the spectrum, and a reinforcement-learning based adaptive one which utilises Q-learning to pick the most appropriate frequency band based on the current state of the spectrum. Both simulated and real-world tests with a doppler radar are carried out to assess which approach is more successful, and to determine appropriate values for the learning rate and discount factor in the reinforcement learning-based approach. Tests show that the reinforcement learning approach is more successful both in not being jammed, and (for the adaptive solutions) in transmitting on the same frequency for longer, which enables detection of slow targets.