Multi-stage Global Sensor Fusion for Navigation using Nonlinear Observers and eXogeneous Kalman Filter

Reduced cost, miniaturization, and increased availability of advanced sensors systems is a major driving force and enabler for new technology, products and services. GPS, cameras, inertial measurement units, pressure sensors, radio/network positioning, and magnetic sensors can now be embedded into almost any device. At the core of these systems there is advanced software that translates the measurements into accurate and reliable estimates of position, velocity and attitude. This processing depends on mathematical models of the sensor systems and the user (e.g. vehicle motion), and is called navigation sensor fusion. To ensure reliability and performance of emerging applications, the project focuses on more robust and reliable processing architectures, algorithms and software for navigation sensor fusion. The project addresses fundamental properties of a new theoretical basis and will develop an open source software environment for rapid and efficient implementation of navigation applications. Experimental demonstrators have targeted emerging applications of drones that must operate under strict requirements for navigation accuracy and robustness in order to ensure the success and safety of missions. This includes operating conditions when satellite navigation is not available, such as industrial inspection close to structures that are blocking or disturbing the satellite signals and in poor visibility.

The knowledge that is established through this project is of fundamental nature, and it forms a foundation for further theoretical and application-oriented research. This concerns in particular results on global stability and robustness of SLAM and new results on Kalman filtering. The software for state estimation for sensor fusion and navigation is advanced and with high TRL. It is a platform for applied research and has a potential for commercial exploitation. Postdoc and doctoral candidates that were financed by the project have completed successfully and entered into new positions in academic, research and industry.

Reduced cost, miniaturization, and increased availability of advanced sensors systems is a major driving force and enabler for new technology, products and services. GPS, cameras, inertial measurement units, pressure sensors, radio/network positioning, and magnetic sensors can now be embedded into almost any device. At the core of these systems there is advanced software that translates the measurements into accurate and reliable estimates of position, velocity and attitude. This processing depends on mathematical models of the sensor systems and the user (e.g. vehicle motion), and is called navigation sensor fusion. To ensure safety, reliability and performance of emerging applications, the project focuses on more robust and reliable processing architectures, algorithms and software for navigation sensor fusion. A main hypothesis is that the navigation sensor fusion problem can be solved with high accuracy, robustness and reliability using estimators and observers that have individual strong global properties. The underlying theoretical platform for the project is the use of estimator modules having strong global stability properties, in cascaded and weak feedback interconnections that ensures that the total system inherits their strong global stability properties. The project funds two postdoctoral researchers and one doctoral fellow, and is organized in 3 work packages: WP1: Theoretical basis for multi-stage navigation filtering, including a performance-enchancing use of the Kalman-filter where a linear time-varying model is computed by an exogenous nonlinear observer, the eXogenous Kalman Filter (XFK). WP2: Globally stable navigation sensor fusion using nonlinear range and range-rate measurements, including simultaneous localization and mapping. WP3: Open source navigation software based on Matlab GNC toolbox and automatically generated embedded C code. Experimental demonstrators targets emerging applications of unmanned aerial vehicles.