Execute-Evaluate Two-Stage Framework for Intelligent Jamming Decision-Making Based on Reinforcement Learning

With the rapid development of cognitive radar, its antijamming capabilities have continuously improved, posing a challenge to the decision-making capabilities of jammers. To generate effective jamming strategies, an intelligent jamming strategy learning method is proposed. In our formulation, the jamming performance evaluation is of vital importance and an online jamming effect evaluation mechanism is first established, which divides the jamming process into two stages: 1) execution and 2) evaluation. In addition, the concept of minimum effective jamming power spectral density is introduced to help the jammer determine the appropriate power levels and provide auxiliary information for decision-making. Building on these properties, the interaction between the jammer and the radar is modeled as a Markov decision process, where the jammer acts as the agent and the radar constitutes part of the environment. Through continuous interaction between both sides, the optimal jamming types and transmit/receive modes are explored based on the proximal policy optimization algorithm. Simulation results demonstrate that the jammer can learn optimal jamming strategies for various radar transmission strategies without relying on the radar's internal performance indicators.