A Jamming Decision-Making Method Based on Offline Reinforcement Learning

Electronic warfare (EW) plays a critical role on modern battlefields, with its core being the disruption and deception of enemy electronic equipment to weaken their combat capabilities. Firstly, the electromagnetic environment of modern battlefields is complex and variable, and enemy targets may possess certain anti-jamming measures. It is challenging for the jamming party to effectively interfere with the targets using inherent strategies. Secondly, testing real-time jamming strategies may involve complex and highly sensitive environments, making it difficult to analyze the performance of the obtained strategies before actual deployment, and challenging to ensure their effectiveness in the operational environment. This paper proposes a method based on offline reinforcement learning. By leveraging abundant historical game sequence data, this method can be trained without direct interaction with the environment, thereby generating a jamming strategy with guaranteed performance. This approach makes the decision-making of the jamming party efficient and safe in complex electromagnetic environments.