Neural Network-based GPS/INS Integrated System for Spacecraft Attitude Determination

Global Positioning System (GPS)/Inertial Navigation System (INS) integrated system is continuously gaining research interests in many positioning and navigation fields. Kalman filtering-based integrated algorithm has some drawbacks on stability, computation load, robustness, and system observability performances. Based on neural network technology, a new GPS/INS integration filtering algorithm is studied for an integration scheme of the attitude determination GPS/INS integrated navigation system. Through some theoretic analysis, this algorithm not only has good estimation performance, but also has better robustness to the system model and noise than the traditional Kalman algorithm. To assess the performance of the proposed integrated model more deeply, some simulation is done to compare with the traditional Kalman filter model. The results indicate that the proposed model provides a significant improvement in some performance, such as accuracy, stability, robustness, and so on.