Multi-Domain Cognitive Anti-Jamming Engine Based on DL and CBR for Multi-UAV Communications

In view of the complex electromagnetic environment, how to reasonably select communication parameters against dynamic jamming has become a research hotspot for multiple unmanned aerial vehicles (UAVs) communication. Based on deep learning (DL) and case-based reasoning (CBR), the decision-making process of the anti-jamming engine designed in this paper is divided into offline and online stages. Offline stage trains convolutional neural network (CNN) with the jamming feature map received by multiple UAVs as the input and the normalized multi-objective function values as the output. At the same time, a case base corresponding to objective function values and decision results is established. During online stage, the real-time jamming feature is input into the trained network to obtain the function value, and then the decision result is obtained according to the case base corresponding to the function value. In this paper, real-time decision-making is achieved in a dynamic environment by changing the jamming pattern.