Low-cost failure-tolerant hybrid navigation designs for future space transportation systems

Abstract

Global Navigation Satellite System (GNSS) technology is steadily making its way into space transportation systems. However, the most widely used launcher navigation systems to date, including those on board of European launch vehicles, are inertial-only. Despite reliable and robust, the dead-reckoning nature of these systems causes unbounded navigation solution drift, raising important mission constraints and limitations. Due to its bounded error characteristics, GNSS has long been used as remedy for such inertial drift. Combined, the two systems compensate many of each others’ shortcomings.

This thesis presents the research and development of a robust and fault-tolerant GNSS/inertial hybrid navigation system for launchers based on lower grade sensory. A comprehensive survey of architectural elements of the GNSS/inertial fusion under the scope of the envisaged application is first carried out, providing the background and justification for the proposed system concept. The design process is broken down into three stages: baseline system concept, robust order reduction, and Fault Detection, Isolation and Recovery (FDIR) scheme design. A tightly-coupled, modular, closed-loop architecture featuring an error-state Kalman filter updated by GNSS pseudorange and time-differenced carrier phase measurements forms the baseline design. The filter runs in parallel with a strapdown inertial propagator, estimating a wide set of inertial sensor perturbations, including many associated with low-grade units, which is fed back to the propagation algorithm. GNSS receiver clock and atmospheric offsets are also mitigated within the filter. Given the large state size, filter order is then reduced through Consider state Kalman filtering. Rederiving this filtering framework, a novel perspective is drawn whereby the standard- and consider-state routines are entirely separated, enabling higher implementation flexibility. Correlation between process and measurement noise is also accounted for. Different inertial sensor filter models (high, medium, and low grade) are reduced based on extensive parametric observability and error impact analyzes using Parametric Cramér-Rao Bounds along representative launch vehicle trajectories. GNSS filter state set is also reduced. Despite using only around half of the computational load of the original full-order estimator, the resulting lower-order filter configurations maintain the robustness of the baseline design against inertial and GNSS sensor errors. Going a step further and targeting common failure modes of hybrid navigation, the developed system is suited with an FDIR module. The proposed novel scheme uses both filter innovation and GNSS measurement-set statistics to detect and isolate GNSS failures as well as inertial/strapdown faults. Threshold selection and concept validation is done through stochastic analysis of the statistical test variables under real GNSS receiver data. Required computation load is shown to be small relative to that of the filter update step.

In a stride to raise the readiness of the designed system the navigation algorithm is implemented as C/C++ software, based on DLR’s Hybrid Navigation System (HNS) heritage, flight-proven code. In this process new navigation software libraries are created and several original ones updated. Despite all the additional performance, robustness, fault-tolerance enhancing features with respect to the heritage software, computational load analysis shows only a moderate burden increase, which does not considerably strain the real-time load margins of the latter.

At each design stage, performance is evaluated through model-in-the-loop Monte Carlo and/or hardware-in-the-loop testing using real GNSS data from a representative receiver stimulated with trajectories of both DLR SHEFEX-2 sounding rocket and ESA Vega launcher. The C/C++ navigation software is tested in a software- and hardware-in-the-loop set-up. The proposed system is compared to several other configurations. Among these, comparison to a representative launcher (high grade) inertial-only system suggests the possibility of a one-grade reduction of the inertial sensor with the conceived hybrid design, significantly lowering system cost.