Intelligent maneuver decision-making for UAVs using the TD3-LSTM reinforcement learning algorithm under uncertain...

Area of Science:

• Artificial Intelligence
• Robotics
• Aerospace Engineering

Background:

• Unmanned aerial vehicles (UAVs) face complex decision-making challenges in aerial confrontations.
• Existing methods struggle with the uncertainty inherent in these scenarios.

Purpose of the Study:

• To develop an intelligent maneuver decision-making method for UAVs in aerial confrontation.
• To address the complexity and uncertainty in UAV combat scenarios.

Main Methods:

• A twin delayed deep deterministic policy gradient (TD3)-long short-term memory (LSTM) reinforcement learning approach.
• A victory/defeat adjudication model based on UAV operational capability and a 3-DOF UAV model.
• A model-driven state transition update mechanism for continuous action space decision-making.
• Uncertainty estimation using Wasserstein distance and memory nominal distribution for target detection noise.

Main Results:

• The proposed TD3-LSTM method effectively extracts features from high-dimensional, uncertain aerial confrontation situations.
• Simulation experiments demonstrate the method's capability in assisting UAVs with maneuvering decisions.
• The system successfully navigates complex aerial combat scenarios.

Conclusions:

• The developed reinforcement learning method enhances UAV maneuver decision-making under uncertainty.
• This approach offers a robust solution for intelligent autonomous aerial combat.
• Further validation through diverse simulation experiments confirms the method's effectiveness.