Approximating Nash equilibrium for anti-UAV jamming Markov game using a novel event-triggered multi-agent...

Area of Science:

• Wireless Communications
• Artificial Intelligence
• Game Theory

Background:

• Ground users face challenges from intelligent aerial jammers in downlink communication.
• The interaction between ground users and jamming UAVs forms a Markov game for anti-jamming.
• Existing multi-agent reinforcement learning (MARL) methods struggle with high dimensionality and local optima.

Purpose of the Study:

• To propose a novel algorithm, ETMAPPO, addressing MARL limitations in anti-UAV jamming scenarios.
• To reduce information transmission dimensionality and enhance policy convergence efficiency.
• To achieve stable jamming and anti-jamming strategies approximating a Nash equilibrium.

Main Methods:

• Developed an event-triggered mechanism for adaptive information observation by agents.
• Incorporated a Beta operator to expand the policy space search for optimized actions.
• Utilized multi-agent reinforcement learning (MARL) within a model-free framework.

Main Results:

• The proposed ETMAPPO algorithm demonstrated superior global benefits with reduced information dimensionality compared to benchmarks.
• Ablation studies confirmed the effectiveness of the event-triggering and Beta strategy components.
• Convergence performance indicated the algorithm's capability to establish stable jamming and anti-jamming strategies.

Conclusions:

• ETMAPPO offers an efficient solution for the anti-UAV jamming Markov game.
• The algorithm effectively mitigates challenges of dimension explosion and local convergence in MARL.
• The developed strategies approximate a Nash equilibrium, enhancing communication security against aerial threats.