Neural network assisted inverse dynamic guidance for terminally constrained entry flight.

Area of Science:

• Aerospace Engineering
• Control Systems
• Artificial Intelligence

Background:

• Spacecraft reentry requires precise trajectory control to meet terminal constraints.
• Traditional guidance laws face challenges in achieving fully constrained flight, especially terminal velocity.

Purpose of the Study:

• To develop a novel neural network-assisted entry guidance law.
• To ensure full satisfaction of terminal constraints including position, flight path, azimuth, and velocity.
• To improve the efficiency and performance of entry flight control.

Main Methods:

• Utilizing Bézier approximation for fully constrained trajectory approximation.
• Solving the inverse dynamic system for entry flight to generate guidance commands.
• Employing an artificial neural network for terminal velocity prediction and Bézier curve adjustment.

Main Results:

• The Bézier approximation successfully models reference trajectories.
• The guidance law ensures terminal constraints for position, flight path, and azimuth angle.
• Neural network prediction of terminal velocity allows for curve adjustment, meeting velocity constraints.
• Simulations demonstrate improved performance of the neural network-assisted method.

Conclusions:

• The proposed method offers a faster implementation for fully constrained entry flight.
• The neural network-assisted guidance law shows significant performance improvements.
• The scheme holds promise for automated onboard control in reentry and terminal guidance.