Precision Attitude Determination Using a Multiple Model Adaptive Estimation Scheme

This paper is mainly motivated by three reasons: (1) future missions which will necessitate the employment of low cost and low grade Micro-Electro-Mechanical Systems (MEMS) sensors (e.g., MEMS gyros or compact star trackers) while still demanding a high precision attitude estimation, (2) development of a real-time noise statistics estimation capability in order to extend/enhance the performance of a traditional Kalman estimator whose performance is mainly dictated by the knowledge accuracy of its process noise and measurement noise covariance matrices, and (3) performance enhancement of a traditional 6 state Extended Kalman Filter (EKF) whose performance is drastically affected and compromised due to its inability to account for scale factor (SF) errors and misalignment errors. Three specific design areas to be investigated in this paper include: (1) the design and implementation of an attitude determination system (ADS) using a Multiple Model Adaptive Estimation (MMAE) scheme wherein the mixing of various EKF models reflecting various state dimensions is employed to accommodate for SF errors and misalignment errors at high rate operating conditions, (2) real-time gyro noise statistics (rate random walk, angular random walk, and SF errors) estimation via an additional MMAE scheme implemented in parallel to provide process noise update to the ADS individual EKF, and (3) the applicability of MMAE scheme to multi-sensor data fusion for an effective measurement update. The feasibility of the proposed concept and its performance improvement as compared to a traditional approach are evaluated via simulation.