Reinforcement-Learning-Based Agile Transmission Strategy for Networked Radar System Anti-Jamming

Radars employing agile waveforms can enhance detection performance by leveraging parameter agility under jamming environment. To further exploit and augment the detection capabilities of networked radars, this paper introduces an agile transmission strategy (ATS). This strategy enables radars to dynamically choose between detection and covert operation, improving anti-jamming capabilities while maintaining robust detection performance. To optimize the transmission configurations of radar nodes, we propose a reinforcement learning based Markov decision process model. This approach enhances pulse accumulation efficiency and anti-jamming performance by learning the jammer's strategy through interaction. Given the nonlinear relationship between detection probability and pulse accumulation, we use detection probability, rather than the signal-to-interference-plus-noise ratio (SINR), as the reward function. This choice allows for a more precise alignment with detection requirements. To minimize training time and support real-time online application, Q-learning is utilized to derive the optimal strategy in electronic countermeasure environments. In the scenario considered, multiple radars are mounted on mobile platforms and intermittently emit pulses to reduce the probability of interception and mitigate the risk of being targeted. Real-time anti-jamming strategies are developed during the electronic countermeasure process. Simulation results confirm that the proposed method effectively reduces radar jamming probability while enhancing detection performance.