Celestial Aided Inertial Navigation by Tracking High Altitude Vehicles

Celestial and inertial navigation systems work well together as an alternative to Global Positioning System for navigation. The inertial system provides constant dead reckoning navigation updates but is subject to drift errors. Celestial systems then limit the inertial errors with its passive stellar measurements. Celestial measurements normally provides attitude updates only by tracking the angular positions of known stars and determining a deviation from the estimated vehicle attitude. However, by imaging reference objects with known positions and velocities against a background field of stars, position and velocity can be determined. With the ubiquitous use of aircraft, cooperative vehicles can be a ready source of information from which to make a reference, assuming that the observer can get continuous updates about the reference vehicle’s position via communication links. This research uses a MATLAB tool developed by the Air Force Research Laboratory that simulates a remotely piloted vehicle with celestial and inertial sensors and a barometric altimeter. The vehicle tracks a higher altitude aircraft with its camera to aid the inertial system. Three different scenarios are studied: 1) stellar observations providing attitude updates only, 2) aircraft observations providing bearing measurements to known position and velocities, and 3) both stellar and aircraft observations. Additionally, the observation frequency will be a variable parameter to determine its effect on navigation accuracies. The sensor measurements are combined using an extended Kalman filter. A control scenario showing the result of dead reckoning only is also performed for comparison. As expected, scenario 1 shows poor navigation aiding. Scenario 2 provides a significant improvement in position and velocity aiding while having comparable attitude accuracy. Scenario 3 provides the same level of position and velocity accuracy as scenario 2, but greatly improves the attitude performance. The results also show that the frequency of measurement updates has an insignificant impact to aiding in any scenario.