Abstract: As one of the most important tools for space and aerial information collection, microelectromechanical systems (MEMS) sensors require real-time and rapid detection and correction of their time/space coupling errors to ensure the security of sensor information and the reliability of data processing. Among the existing sensor error detection and correction algorithms, the methods based on traditional Kalman filtering (KF) have good real-time performance and low computational cost, but lack robustness in the presence of nonlinearity, non-Gaussian noise, and strong signal interference. Although the methods based on deep learning have better error detection and correction effects, they suffer from high computational cost and poor real-time performance. To address these issues, this paper innovatively combines the above two types of methods and proposes an intelligent detection and correction algorithm for time/space coupling errors of space and aerial MEMS sensors based on a dual dynamic saturation mechanism. Firstly, by leveraging the idea of KF filtering, an innovation saturation-based error detection algorithm is designed to dynamically detect errors. Secondly, a genetic algorithm is introduced to adaptively optimize the nonlinear parameters in the core formula for determining the dynamic saturation threshold, thereby enhancing the algorithm's generalization performance across different datasets. Then, to address the low optimization efficiency and high number of iterations of the genetic algorithm, a novel dual dynamic saturation mechanism is proposed to reduce the time/space coupling relationship between nonlinear parameters, thereby improving the parameter optimization efficiency and accelerating convergence. Finally, an experimental platform is built to collect MEMS sensor data. The algorithm verification based on the measured data shows that, compared with the traditional KF detection framework and neural network detection framework, the proposed intelligent error detection and correction algorithm has significant performance improvements in terms of detection and correction efficiency and accuracy.

Key words: Error detection and correction, Deep learning, Kalman filtering, MEMS sensors, Adaptive optimization