Priori-assisted soft actor–critic based interrupted sampling repeater jamming method

Interrupted sampling repeater jamming (ISRJ) is an electronic countermeasure widely used in airborne self-defense jamming equipment, with its performance heavily dependent on the jamming parameters. In non-cooperative and dynamic countermeasure scenarios, existing jamming decision-making methods that rely heavily on expert knowledge bases and precise mathematical models are prone to failure, making it challenging to achieve intelligent and adaptive jamming. Furthermore, due to the multi-dimensional and continuous nature of parameter spaces, these methods tend to converge slowly. To address these issues, we propose a priori-assisted soft actor–critic (SAC) based ISRJ method. In our method, we first model the ISRJ decisions throughout the entire penetration process as a Markov decision process (MDP) and design distinct reward functions for both cooperative and non-cooperative scenarios. We then utilize the SAC algorithm to learn the optimal strategy in an unknown, dynamic environment while efficiently managing large state–action spaces. To accelerate convergence while effectively avoiding potential local optima, we propose the priori-assisted SAC algorithm to solve the above MDP. Compared to the classical SAC, the proposed algorithm integrates expert prior information, which assists the agent in exploring more effectively, thereby improving both the efficiency and quality of policy learning. Extensive simulation results confirm the superiority of the proposed method.