MADDPG-Based Collaborative Anti-Jamming Strategy for Joint Frequency-Power Allocation in Networked Radars

In modern electronic warfare, main-lobe jamming in complex environments poses a significant threat to radar system survivability. To address this challenge, this paper proposes a collaborative anti-jamming strategy generation algorithm for networked radar systems, based on the Multi-Agent Deep Deterministic Policy Gradient framework. The radar-jammer interaction is modeled as a Partially Observable Markov Decision Process, and a dual-signal transmission framework for deception and detection is designed to mislead jammers into attacking the deception signal, thereby enhancing the anti-jamming capability. By leveraging multi-agent reinforcement learning, the proposed method dynamically optimizes joint frequency–power allocation in complex environments. Simulation results demonstrate that the proposed approach effectively mitigates dynamic main-lobe jamming and achieves superior stability and robustness in anti-jamming performance.