Enhancing multi-UAV air combat decision making via hierarchical reinforcement learning.

Area of Science:

• Robotics and Artificial Intelligence
• Aerospace Engineering
• Computational Intelligence

Background:

• Autonomous decision-making is crucial for Unmanned Aerial Vehicle (UAV) air combat.
• Current rule-based algorithms struggle with complex multi-UAV combat scenarios.
• Optimizing autonomous systems in dynamic combat environments remains a significant challenge.

Purpose of the Study:

• To propose a novel hierarchical reinforcement learning (HRL) approach for multi-UAV air combat decision-making.
• To address the limitations of existing methods in complex combat environments.
• To improve the efficiency and effectiveness of autonomous UAV tactics.

Main Methods:

• Designed a hierarchical decision-making network based on tactical action types to simplify maneuver selection.
• Decomposed high-quality combat experience to increase valuable training data and ease strategy learning.
• Validated the algorithm's performance using the JSBSim UAV simulation platform.

Main Results:

• The proposed HRL algorithm demonstrated superior performance compared to baseline methods.
• Effective decision-making was achieved in both even and disadvantaged air combat scenarios.
• The method successfully streamlined the decision-making space and enhanced strategy learning.

Conclusions:

• The novel hierarchical reinforcement learning approach offers a significant advancement in multi-UAV air combat.
• This method provides a more effective solution for complex autonomous decision-making in aerial warfare.
• The findings suggest a promising direction for future research in intelligent UAV systems.