Cognitive Jamming Policy Generation Based on A2C Algorithm

Due to the rapid development of cognitive electronic warfare (CEW) and complex electromagnetic environments, traditional jamming methods are difficult to adapt to new types of scenarios. Reinforcement learning (RL), as one of the cores of machine learning, provides a solution for jammers to adapt to new challenges. In order to quickly find a jamming policy for multifunctional radar (MFR), this paper proposes a jamming policy generation scheme based on the Advantage Actor Critic (A2C) algorithm, which utilizes interactive self-learning to dynamically adjust the jamming policy and realize cognitive jamming decision generation. Firstly, this paper investigates the model of jamming policy generation and discusses the relationship between radar operating modes and jamming modes, and the whole process can be modeled as a Markov decision process (MDP). In this paper, a heuristic reward function is used, which helps the jammer to better explore the optimal strategy. Secondly, we discuss the A2C algorithm, especially the advantages of this algorithm over other RL methods for which research results are available. Finally, in this paper, we compare the simulation results of the A2C algorithm with Q-Learning and DQN algorithms, and demonstrate that the algorithm effectively improves the cognitive jamming policy generation ability of the jammer.