Scorpion: A Modular Sensor Fusion Approach for Complementary Navigation Sensors

There is a great need to decrease our reliance on GPS by utilizing novel complementary navigation sensors. While a number of complementary navigation sensors have been studied, each one has trade-offs in availability, reliability, accuracy and applicability in various environments. The development of a robust estimator therefore requires the integration of many diverse sensors into a sensor fusion platform. Unfortunately, as the number of sensors added to the system grows larger, so does the difficulty of developing a sensor fusion solution that optimally integrates them all into a single navigation estimate. In addition, a sensor fusion solution with many sensors is susceptible to sensor failures, modeling errors, and other phenomena which can cause degradation of the fusion solution. In this paper, we propose an open architecture for sensor fusion that allows for the development of modular navigation filters, sensor integration strategies, and integrity algorithms. The primary goal of this architecture is to allow for the rapid development of a novel complementary PNT sensor, fusion strategy, or integrity algorithm without modification of any other part of the system. In the future, this architecture will enable the community to develop a repository of well-tested software modules for sensor fusion which will in turn allow for the iterative development of robust estimators, where users may pick and choose the components that they wish to use from the repository and build an estimator that fits their application. In addition, domain experts in the community on a particular sensor phenomenology may contribute modules to the repository without needing to be experts in all aspects of sensor fusion. To facilitate this community engagement, we have developed an open source implementation of the architecture, which will be made available as a reference implementation of the architecture and approach. This paper details the design and overall approach to the open architecture, as well as shows some experimental results that were obtained by running flight data through the reference implementation.