Optimal Estimation of Dynamic Ionosphere Induced Group Delays of GPS Signals

J. Arnold Soltz and Andrea M. Johnson

Abstract:	There are many sources of navigation error in a received Global Positioning System (GPS) satellite signal. Among these is the ionospheric group delay. In the past, singlestate, dual-frequency filters have been used to estimate the delay for authorized users. Although sufficient for terrestrial receivers, such a filter is inadequate for many space-based missions. Recently, Charles Stark Draper Laboratory designed a real time five-state Kalman filter to be used on re-entering space vehicles and has modified a GPS receiver to implement the algorithm. Three measurements can be used as input to this filter: dual-frequency pseudo-range differencing and single-frequency rate measurements for each link frequency (L1 and L2). The single-frequency rate measurements make use of the relationship between carrier phase advance and group delay to provide information about the rate of change of the group delay. The five-state filter models the ionospheric delay dynamics as the third integral of white noise (3 states). An additional Kalman delay state is required to use each of the rate measurements. The dynamics model was tuned for expected gaps in measurements up to 30 seconds; however, because of the simplified dynamics model, a careful reset of states and covariances is necessary if measurements recur after a long absence. The modified GPS receiver performed well when tested in a re-entry scenario implemented in an RF satellite signal simulator. Continued work resulted in design of a seven-state filter (not yet implemented in the GPS receiver) that would improve performance for some mission scenarios. This new filter utilizes four measurements: dual-frequency pseudo-range differencing, dual-frequency delta-range differencing, and single-frequency rate measurements for both frequencies (L1 and L2). Two states are necessary for the model dynamics plus five delay states for processing rate measurements. Instead of performing a separate reset when experiencing measurement loss, the reset is incorporated into the state transition matrix and accomplished at each extrapolation. The process model selected for the seven-state filter was the integral of a first-order Markov process. The filter was used to estimate both the whole ionospheric group delay and the deviation of the delay from a given model. When used to estimate the deviation of the delay from a model, the group delay transitioned from “estimated” to “modeled” smoothly in the absence of measurements. In the absence of measurements, the estimated group delay tends to a bias from the model provided.
Published in:	Proceedings of the 63rd Annual Meeting of The Institute of Navigation (2007) April 23 - 25, 2007 Royal Sonesta Hotel Cambridge, MA
Pages:	485 - 495
Cite this article:	Soltz, J. Arnold, Johnson, Andrea M., "Optimal Estimation of Dynamic Ionosphere Induced Group Delays of GPS Signals," Proceedings of the 63rd Annual Meeting of The Institute of Navigation (2007), Cambridge, MA, April 2007, pp. 485-495.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In