A Deep Reinforcement Learning-Based Shortwave Multistation Autonomous Cooperative Direction Finding and Localization Method

Abstract

Multistation direction finding and passive localization have a wide range of applications in the fields of information countermeasures and navigation. However, signal detection and recognition based on deep learning require a lot of manual annotation, and the localization algorithm requires a large amount of computation, resulting in poor shortwave signal direction finding and localization in real time. To solve the time-consuming and labor-intensive problem, we propose a shortwave direction finding and localization method based on deep reinforcement learning (DRL). The direction of arrival (DOA) of the source signal can be used to locate the shortwave source, but the result of direction finding is not ideal in the case of the same frequency interference. In the proposed method, a space-time high-resolution processing method is designed to improve the accuracy of obtaining directional results under the complex shortwave background. The Cramér-Rao lower bound (CRLB)-based reward function is also designed, and a Markov decision process (MDP) is established. The autonomous direction finding and localization environment is trained by deep Q-network (DQN) and double-DQN (DDQN), and the comparison shows that the DDQN performs better compared with DQN. Finally, we save the online model and compare its positioning performance with other classical algorithms. The simulation experiment results and performance tests verify the feasibility of the proposed method.