Calibration Method for Relativistic Navigation System Using Parallel Q-Learning Extended Kalman Filter.

Area of Science:

• Spacecraft navigation
• Relativistic effects in astrodynamics
• Estimation theory

Background:

• Relativistic navigation systems rely on precise measurements of phenomena like stellar aberration and gravitational light deflection.
• Inter-star angle measurement bias, stemming from star sensor misalignment, significantly degrades navigation accuracy.
• Existing methods struggle to effectively calibrate these biases in real-time.

Purpose of the Study:

• To introduce a novel parallel Q-learning extended Kalman filter (PQEKF) for measurement bias calibration in relativistic navigation.
• To enhance the accuracy of spacecraft position and velocity estimation by mitigating inter-star angle measurement errors.
• To automatically tune the filter's process noise covariance matrix using Q-learning.

Main Methods:

• Development of a parallel Q-learning extended Kalman filter (PQEKF) algorithm.
• Integration of Q-learning for adaptive tuning of the process noise covariance matrix.
• Extraction of relativistic perturbations from inter-star angle measurements using high-accuracy star sensors.

Main Results:

• The PQEKF effectively estimates spacecraft position, velocity, and measurement bias parameters.
• Numerical simulations in a Medium Earth Orbit (MEO) scenario demonstrate the method's high performance.
• Achieved positioning accuracy below 300 meters with inter-star angle measurement accuracy of approximately 1 milliarcsecond (mas) after calibration.

Conclusions:

• The PQEKF offers a robust solution for measurement bias calibration in relativistic navigation.
• The adaptive nature of Q-learning enhances the filter's ability to handle dynamic noise characteristics.
• This approach significantly improves the overall navigation performance and reliability of spacecraft systems.