Energy-efficient strategy generation for smart jammer in non-zero-sum games: A deep reinforcement learning approach

In this paper, we investigate the radio communication countermeasures (RCC) between the smart jammer (SJ) and the intelligent wireless communication system (WCS), focusing on power efficiency. While programmable radio technologies empower SJ with various attack modes, current jamming decision-making methods still primarily target rule-based opponents and operate within the traditional decision framework focused on full-power suppression. These methods are inadequate in addressing the jamming problems involving new degrees of freedom, exhibiting low energy efficiency, and failing to deal with intelligent opponents. In response, we propose a multi-mode intelligent jamming scheme, where the SJ aims to maximize the jamming efficiency by selecting working modes and adjusting jamming power. Firstly, we model the RCC as a non-zero-sum game and analyze the existence conditions of Nash Equilibrium (NE). Then, we propose a proximal policy optimization-clipped (PPO-Clip) jamming strategy generation algorithm based on deep reinforcement learning (DRL) to assist in optimal mode selection and power control, achieving the balance between jamming success rate and power consumption. Finally, we conduct simulation experiments on single-intelligence and heterogeneous “smart versus smart” scenarios. The results indicate that the proposed approach can enhance the overall jamming efficiency and outperform the baseline methods.