A data fusion approach for distributed orbit estimation

To achieve good on-board SAR processing results a precise knowledge of the position of the concerned satellites is essential. For this reason we develop real-time orbit estimation and calibration algorithms which in spite of the small time frame improve the position estimates in an efficient way. Considering future satellite cluster missions (e.g. Cartwheel, Pendulum, Techsat 21, ...), it will be possible to reduce the computing time by splitting of the processing load onto several algorithms that can be implemented in the individual satellites' hardware. The satellites can be considered as a distributed sensor network, which individual (sensor-) nodes have local processors that are used to calculate state estimates using all available data (we want to combine GPS derived and intersatellite distance measurements). A node to node communication is necessary to ensure that no information is lost in order to yield the best possible estimates. The paper specifies advantages and disadvantages of decentralized estimation and compares computational burden and estimation accuracy of decentralized and standard Kalman filter approaches and also analyzes the communication overhead