Reinforcement-Learning-Enhanced Adaption of Signal Power and Modulation for LPI Radar System

Effective implementation of low probability of intercept (LPI) techniques is crucial for enhancing the survivability of radar systems in electronic warfare scenarios. This article explores the use of reinforcement learning in dynamically generating LPI signals in unknown adversarial environments encountered by electronic intelligence (ELINT) systems. We develop a Markov decision model to guide the LPI radar in mitigating power-based interceptions by jointly optimizing transmit signal power and modulation strategies. The interception efficacy of ELINT and the radar's normalized instantaneous transmit power are integrated to comprehensively evaluate LPI performance and radar detection capabilities. A tunable weighting factor facilitates adaptable adjustments between these objectives during decision making. We validate the efficacy of the proposed approach by numerical simulations. Assessment of LPI efficacy is conducted through the analysis of time–frequency modulation signals observed by ELINT, while detection performance is evaluated through synthetic aperture radar imaging tasks.