The design of National Aeronautics and Space Administration Orion's prelaunch navigation system is introduced, both for the first flight test, Exploration Flight Test 1, and for the first planned Orion mission, Exploration Mission 1. A detailed tradeoff of possible design decisions is discussed, and the choices made by Orion are presented. The actual performance of the navigation system during Exploration Flight Test 1 is presented together with the navigation flight-software data provided by Orion to the ground controllers in telemetry.

Launched in December 2014 atop a Delta IV Heavy from the Kennedy Space Center, the Orion vehicle's Exploration Flight Test 1 successfully completed the objective to stress the system by placing the uncrewed vehicle on a high-energy parabolic trajectory, replicating conditions similar to those that would be experienced when returning from an asteroid or a lunar mission. Unique challenges associated with designing the navigation system for Exploration Flight Test 1 are presented with an emphasis on how redundancy and robustness influenced the architecture. Two inertial measurement units, one GPS receiver, and three barometric altimeters comprise the navigation sensor suite. The sensor data are multiplexed, using conventional integration techniques, and the state estimate is refined by the GPS pseudo- and delta-range measurements in an extended Kalman filter that employs UDU factorization. The performance of the navigation system during flight is presented to substantiate the design.

One method to account for parameters errors in the Kalman filter is to ‘consider’ their effect in the so-called Schmidt-Kalman filter. This paper addresses issues that arise when implementing a consider Kalman filter as a real-time, recursive algorithm. A favorite implementation of the Kalman filter as an onboard navigation subsystem is the UDU formulation. A new way to implement a UDU Schmidt-Kalman filter is proposed. The non-optimality of the recursive Schmidt-Kalman filter is also analyzed, and a modified algorithm is proposed to overcome this limitation.

The Orion attitude navigation design is presented, together with justification of the choice of states in the filter and an analysis of the observability of its states while processing star tracker measurements. The analysis shows that when the gyroscope biases and scale factors drift at different rates and are modeled as first-order Gauss-Markov processes, the states are observable so long as the time constants are not the same for both sets of states. In addition, the inertial-measurement-unit-to-star-tracker misalignments are modeled as first-order Gauss-Markov processes and these states are estimated. These results are used to finalize the design of the attitude estimation algorithm and the attitude calibration maneuvers.

Nonlinear filters are often very computationally expensive and usually not suitable for real-time applications. Real-time navigation algorithms are typically based on linear estimators, such as the extended Kalman filter (EKF) and, to a much lesser extent, the unscented Kalman filter. This work proposes a novel nonlinear estimator whose additional computational cost is comparable to (N - 1) EKF updates, where N is the number of recursions, a tuning parameter. The higher N the less the filter relies on the linearization assumption. A second algorithm is proposed with a differential update, which is equivalent to the recursive update as N tends to infinity.

This Note presents a dual accelerometer usage in an orbital Kalman filter. The accelerometer is both used to propagate position and velocity during maneuvers and to update the accelerometer bias state outside of maneuvers. The advantage of this approach is its superior performance to a simple thresholding of the accelerometer. In the simple thresholding scheme the correlation between accelerometer bias and position and velocity during the maneuvers is not sufficient to adequately estimate the bias during coast flight. Therefore the estimate of the bias degrades outside of maneuvers adding considerable uncertainty during subsequent maneuvers. A common solution to this problem is to estimate the accelerometer bias outside of the navigation filter just before a maneuver is performed. A simple averaging scheme is often used, but a Kalman filter is also a possibility. The advantage of the proposed scheme over a stand-alone bias estimator is that a single estimator is globally optimal because it accounts for all the correlations.