Trajectory Tracking Control of Variable Sweep Aircraft Based on Reinforcement Learning.

Area of Science:

• Aerospace Engineering
• Control Systems
• Artificial Intelligence

Background:

• Trajectory tracking is crucial for advanced aircraft maneuvers.
• Uncertainty and environmental changes pose significant challenges to flight control.
• Existing control methods may struggle with the dynamic complexities of morphing aircraft.

Purpose of the Study:

• To develop an advanced control algorithm for trajectory tracking in uncertain environments.
• To enhance the adaptability of flight controllers to dynamic flight conditions.
• To validate the effectiveness of the proposed controller for morphing aircraft.

Main Methods:

• An incremental deep deterministic policy gradient (IDDPG) algorithm was developed.
• The line-of-sight (LOS) method was integrated for path tracking.
• Long short-term memory (LSTM) units were incorporated to improve adaptability.
• A reward function based on position and attitude errors was formulated.

Main Results:

• The IDDPG algorithm demonstrated effective trajectory tracking for a four-wing variable sweep (FWVS) aircraft.
• The controller showed improved adaptability to environmental changes due to LSTM integration.
• Simulations with disturbance factors confirmed the controller's robustness in tracking climbing trajectories.

Conclusions:

• The proposed IDDPG algorithm offers a robust solution for trajectory tracking control of FWVS aircraft under uncertainty.
• Integrating LSTM units significantly enhances the controller's adaptability to dynamic flight conditions.
• The developed controller is effective in managing morphing aircraft during complex maneuvers with external disturbances.