Nathan B. Stastny (AFRL, Space Vehicles Directorate), Frank R. Chavez (AFRL, Space Vehicles Directorate), Chin Lin (AFRL, Space Vehicles Directorate), T. Alan Lovell (AFRL, Space Vehicles Directorate), Robert A. Bettinger (AFRL, Space Vehicles Directorate), James Luck (University of Texas at Austin)
Keywords: Astrodynamics
Abstract:
Since the development of Luigi G. Jacchia’s first density model in 1970 (J70), atmospheric density modeling has steadily focused on large monolithic codes that provide global density coverage. The most recent instantiation of the global density model is the Jacchia-Bowman 2008 (JB08) model developed by Bruce Bowman of the Air Force Space Command. As the models have evolved and improved, their complexity has grown as well. Where the J70 model required 2 indices and various time averages to determine density, the JB08 model requires 5 indices to determine density. Due to computational complexity, the number of real-time inputs required, and limited forecasting abilities, these models are not well suited for onboard satellite orbit propagation.
In contrast to the global models, this paper proposes the development of a density prediction tool that is only concerned with the trajectory of a specific satellite. Since the orbital parameters of most low Earth orbiting satellites remain relatively constant in the short term, there is also minimal variation in the density profile observed by the satellite. Limiting the density model to a smaller orbit regime will also increase the ability to forecast the density along that orbital track. As a first step, this paper evaluates the feasibility of using a localized density prediction algorithm to generate the density profile that will be seen by satellite, allowing for high-accuracy orbit propagation with minimal or no input from the ground.
The algorithm evaluated in this paper is a simple Yule-Walker auto-regressive filter that, given previously measured density values, provides predictions on the upcoming density profile. This first approach requires zero information about the satellite’s current orbit, but does require an onboard method for determining the current, local density. Though this aspect of the onboard system is not analyzed here, it is envisioned that this current, local density (or equivalently drag acceleration) would be calculated through onboard processing of GPS or accelerometer data.
Using the trajectory of the CHAMP satellite as a test case, various samples of CHAMP density data in the past (i.e. before a chosen epoch time) will be input to the filter, and the filter will in turn predict future” density values (i.e. beyond the chosen epoch time). The effectiveness of the filter will be assessed by comparing its predicted density values to true densities in the CHAMP database at those times. The two major design parameters to be investigated for the auto-regressive filter are the appropriate order of the filter and the past data sample (expressed in terms of both time span of the data and sampling interval). Results describe the prediction accuracy of the filter and length of time over which accuracy is maintained for various parameter settings.
Date of Conference: September 1-4. 2009
Track: Astrodynamics