Intelligent anti-jamming decision algorithm for wireless communication under limited channel state information...

Area of Science:

• Wireless Communications
• Artificial Intelligence
• Signal Processing

Background:

• Deep reinforcement learning (DRL) is effective for anti-jamming but often requires complete Channel State Information (CSI).
• Limited CSI presents a significant challenge for DRL-based anti-jamming systems.
• Optimizing exploration rate decay factors in DRL algorithms is complex.

Purpose of the Study:

• To develop an intelligent anti-jamming decision algorithm for wireless systems operating under limited CSI conditions.
• To propose an automatic adjustment algorithm for the exploration rate decay factor.
• To design a Deep Recurrent Q-Network (DRQN) architecture tailored for this scenario.

Main Methods:

• Modeling the anti-jamming problem as a Partially Observable Markov Decision Process (POMDP).
• Developing a DRQN architecture incorporating Long Short-Term Memory (LSTM) networks for temporal feature learning.
• Implementing an automatic adjustment algorithm for the exploration rate decay factor.

Main Results:

• The proposed algorithm achieves near-optimal performance with an automatically adjusted exploration rate decay factor.
• Significant reductions in convergence time: 45% and 32% faster than Double DQN (DDQN) under periodic and intelligent jamming, respectively.
• Improved normalized throughput and superior convergence performance compared to Deep Q-Network (DQN) and Q-Learning (QL).

Conclusions:

• The novel DRQN-based intelligent anti-jamming algorithm effectively addresses the challenge of limited CSI.
• The automatic exploration rate decay factor adjustment enhances decision-making efficiency and performance.
• This approach offers a robust solution for anti-jamming in wireless communications, outperforming traditional DRL methods.