Abstract
To address the weight and power constraints of auto-alignment for strapdown inertial navigation platforms using non-orthogonal gyroscopes, this research establishes a comprehensive error model and proposes a lightweight calibration algorithm based on the Least Squares (LS) method. Validated through ground tests and actual flight data from China’s lunar exploration program, the algorithm achieves high-accuracy on-orbit calibration with significantly lower computational complexity than traditional Extended Kalman Filters (EKF). Furthermore, a logic-based design enables synchronous fault tolerance and calibration for multi-set gyro systems, ensuring mission success during the critical lunar return and reentry phases.
The Challenge: Precision Navigation under Constraints
Returning a spacecraft from the Moon to Earth requires extreme navigational precision to ensure a safe atmospheric reentry. The vehicle’s Strapdown Inertial Navigation System (SINS) must be perfectly calibrated to handle high-speed maneuvers. However, several factors complicate this on-orbit:
- Mechanical and Thermal Shifts: Launch vibrations and extreme temperature changes in space can alter the physical alignment of sensors.
- Non-Orthogonality: To minimize weight, gyroscopes are often mounted in non-standard configurations rather than perfect 90-degree angles. This introduces complex geometric errors that must be modeled mathematically.
- Computational Limits: While the Extended Kalman Filter (EKF) is a standard tool for state estimation, its high-dimensional matrices are computationally expensive. Flight controllers, which must manage redundant systems simultaneously, often lack the processing cycles required for real-time EKF execution.
Methodology: A Lightweight, Fault-Tolerant Approach
The researchers developed a streamlined solution by replacing recursive filtering with a Least Squares (LS) based calibration algorithm.
Error Modeling and LS Implementation
The team first constructed a mathematical model treating gyro axes as arbitrary vectors in space, accounting for non-orthogonal installation angles, scale factor errors, and constant drifts. By performing specific, constant-velocity rotation maneuvers on-orbit, the system collects stable angular velocity data. This allows the complex error equations to be simplified into a linear system, which is then solved efficiently using the LS method.
Redundancy and Fault Tolerance
A key innovation is the dual-layer software architecture designed for multi-system reliability. The system processes data from two independent gyro groups ("Group 1" and "Group 2") to calculate redundant quaternions.
As shown in the architecture above, the system incorporates diagnostic logic to monitor sensor health. If a fault is detected in one group, the system can seamlessly transition to the backup set, ensuring that the calibration and navigation processes remain uninterrupted.
Validation and Mission Results
The algorithm's performance was verified through a combination of 1,000 Monte Carlo simulations, ground-based semi-physical tests using three-axis turntables, and actual flight telemetry from the Chang'e 5-T1 mission.
- Accuracy: Simulations demonstrated that the LS method matches EKF performance, achieving installation angle accuracy better than 0.01° and drift calibration accuracy better than 0.01°/h.
- Efficiency: The computational load was reduced enough to allow the flight computer to process two full IMU sets within a single control cycle—a feat impossible with standard EKF implementations on the same hardware.
- Flight Success: During the actual lunar reentry flight, post-mission analysis revealed that on-orbit calibration reduced pitch errors from approximately 0.12° (using pre-flight ground data) to a remarkable 0.001°.
Conclusion
This research demonstrates that for scripted on-orbit maneuvers, computational efficiency does not have to come at the cost of accuracy. By leveraging a Least Squares approach and robust fault-tolerant logic, the team successfully implemented a high-precision navigation solution for deep-space missions. This methodology provides a blueprint for future lightweight aerospace designs, particularly for missions where processing power and mass are at a premium.
Summary updated