A robust and self-tuning kalman filter for autonomous spacecraft navigation

Most navigation systems currently operated by the National Aeronautics and Space Administration (NASA) and other major space agencies (e.g., European Space Agency (ESA)) are ground-based, and require extensive support to produce accurate results. Recently developed systems that use Kalman filter and the Global Positioning System (GPS) or GPS-like data for orbit determination greatly reduce dependency on ground support, and have potential to provide significant economies for spacecraft navigation. Current techniques of Kalman filtering, however, still rely on manual tuning from analysts, and cannot help in optimizing navigation autonomy without compromising accuracy and performance. The re-tuning process is more complex when dealing with geosynchronous and high-eccentricity orbits. This dissertation presents an approach to produce a high accuracy onboard navigation system that can recover from estimation errors in real time. It proposes a sophisticated application of neuro-fuzzy techniques to perform the self-tuning capability. It also demonstrates the feasibility and efficiency of a self-tuning component built from this concept to augment to a Kalman filter, which performs the state estimation. The core requirement is a method of state estimation that handles uncertainties robustly, is capable of identifying estimation problems, flexible enough to make decisions and adjustments to recover from these problems, and compact enough to run on flight software. The scope of the dissertation research has both theoretical and experimental dimensions. In the direction of theory, performance limits of Kalman filter and related major adaptive techniques, and new technologies popular in the areas of system identification and automatic controls are studied, with special emphasis on mathematical issues leading to the optimization of spacecraft navigation autonomy. In the experimental direction, a prototype self-tuning system is designed, developed, and tested. Filtered data from real and simulated GPS measurements are carefully prepared to train and check the accuracy of the system. The experimental implementation establishes the reliability and accuracy of the mathematical foundations of neuro-fuzzy techniques underlying the self-tuning process. Results from the testing of the prototype show that this self-tuning technique can achieve the accuracy of less than 5 cm in total position. This concept of a robust and self-tuning Kalman filter for autonomous spacecraft navigation is also extended to broaden its mission scope to include geosynchronous orbits and near-Earth high-eccentricity orbits.