FlexiRL: An Integrated Jamming Strategy Framework for Hybrid Mode and Parameter Decision-Making

Electronic warfare holds increasing importance in modern military operations. Effective jamming decision-making is capable of supplying timely electronic protection for high-value targets. Developing a powerful jamming strategy requires selecting appropriate jamming modes and optimizing jamming parameters. To achieve simultaneous optimization of jamming modes and multidomain parameters, a jamming strategy framework based on deep reinforcement learning (RL) is proposed in this article. First, the hybrid mode and parameter decision-making (HMPDM) problem is modeled as an augmented jamming Markov decision process (AJMDP). Especially, a contextual exploration policy is employed in jamming mode selection and hierarchical rewards are well-designed, facilitating quick discovery of effective jamming actions. A state-residual time-varying weighting method is proposed for jamming evaluation to accommodate dynamic environments. Further, a jamming strategy framework of flexible RL is designed to solve the AJMDP model, providing an intelligent solution to the HMPDM problem. It enables dynamic optimization of discrete jamming modes and hybrid discrete-continuous parameters based on the online jamming evaluation. Simulation results demonstrate that the proposed method can learn better policies and realize effective jamming both in time, frequency, and energy domains. Moreover, the proposed method exhibits a superior jamming success rate and needs fewer timesteps to achieve successful jamming compared with prevailing methods.