The fusion of GNSS and INS (Inertial Navigation System) is a well-known method to improve the robustness and availability of the whole system. Due to the complementary features and error sources of the two systems, they supplement each other perfectly.
In the literature different levels of INS/GNSS integration are proposed. They differ mainly on their depth of integration and hence, result in different advantages and disadvantages. We define three classes with respect to the fusion level namely loosely coupled (in the position domain), tightly coupled (on pseudorange level) and ultra-tightly coupled system (feedback on the tracking loops of the GNSS receiver).
The fusion algorithm itself is normally realized by an advanced Kalman filter (extended or unscented Kalman filter) or particle filter taking into account the given sensor error characteristics, nonlinearity of the position solution and the user dynamics.
A potential drawback of very precise inertial sensors is their high cost. Thus, the use of low cost Micro-Electro-Mechanical Systems (MEMS) sensors for civil aviation is a particularly attractive area of investigation at the DLR Institute of Communications and Navigation. GNSS receivers can also provide attitude information by using additional antennas separated by constant baselines. This measurement can be combined with the attitude given by an inertial measurement unit (IMU) to improve robustness.
Airborne Autonomous Integrity Monitoring (AAIM) is a novel concept that provides a measure of integrity for the GNSS/INS solution. Currently, DLR investigates AAIM methods within the frame of the TOtal Performance concept for GBAS-based Automatic Landings (TOPGAL) project. Therefore, the single sensor error characteristics must be well-known and understood as well as their propagation through the fusion algorithm.
For a first insight into inertial related integrity concept, we looked into a rather low-complex fusion algorithm where the INS errors are frequently updated whenever a trustworthy GNSS solution is available. If no suitable GNSS solution is available we over-bound the inertial coasting error depending on the error model of the sensor. The corresponding error parameters can be found via static and dynamic sensor tests.
The figure below shows the error bounds for three different IMU grades, where the low-cost grade is drifting faster and stronger and therefore, results in the larger error performance.