Systems Engineering and Electronics ›› 2018, Vol. 40 ›› Issue (3): 518-525.doi: 10.3969/j.issn.1001-506X.2018.03.05

Previous Articles     Next Articles

Jamming strategy learning based on positive reinforcement learning and orthogonal decomposition

ZHUANSUN Shaoshuai1,2, YANG Junan1,2, LIU Hui1,2, HUANG Keju1,2   

  1. 1. College of Electronic Countermeasure, National University of Defense Technology, Hefei 230037, China; 2. Key Laboratory of Electronic Restriction, Hefei 230037, China
  • Online:2018-02-26 Published:2018-02-24

PDF (PC)

Abstract:

As a self-improving and online learning method, reinforcement learning gets an optimal strategy through trial-and-error by interacting with the dynamic environment continually, and has become an important subfield of machine learning. Current study of the jamming strategy needs priori information or has a slow learning speed, an algorithm of learning jamming strategy based on positive reinforcement learning-orthogonal decomposition (PRL-OD) is proposed. This algorithm increases the possibility of choosing the optimal action by positive reinforcement learning, and then accelerates the learning speed which is also known as the convergence rate. Specially, when signal constellation is distorted by many factors, such as channel noise or imbalance between in-phase and quadrature channels, the proposed signaling scheme orthogonal decomposition method could learn the optimal in-phase component and quadrature component to form the optimal jamming signal. Numerous results show that the proposed PRL-OD algorithm could learn better jamming parameters and faster modulation patterns than state-of-art algorithms. In the same jamming tasks, the PRL-OD algorithm only needs fewer interactions and achieves higher jamming performance.

[an error occurred while processing this directive]