ESA GNC Conference Papers Repository

Future observation missions from the geostationary orbit require extremely accurate geolocation of acquired images. For instance, weather imaging missions need pixels to be located on the Earth within half the pixel size. With a resolution of 500m, this represents a knowledge of the line-of-sight better than 7 ?rad RMS. This leads to stringent requirements in terms of attitude and orbital position knowledge. Starting with internal studies at Astrium in 2006, followed by an R&D study conducted with Eumetsat in 2007, we have designed a family of navigation filters which hybridize measurements from attitude sensors with image data (ground control points). We have proved that such a filter could accurately estimate both attitude and orbital position errors, so that requirements for ground-based orbit tracking could be greatly relaxed. In this paper, we present the general context and the physics of the estimation problem. We detail the structure of the Kalman filter, with the state representation involving both attitude and orbital errors, as well as the observation principles. We then show how a noncausal formulation of the filter can make the performance robust to long gaps between ground control points. We validate the resulting performance both with a covariance analysis and representative time-simulation. Finally, we describe a preferred operational setup recommended with Eumetsat, in which occasional estimates from such a filter are optimally combined with classical ground tracking systems for cost-effective accurate tracking suited to the geolocation needs.