A comparison of nonlinear filters for orbit determination

Daryl G. Boden, Bruce A. Conway

Research output: Contribution to conferencePaper

Abstract

This paper compares the performance of three filters which have been applied to the problem of orbit determination using actual satellite tracking data obtained from ground based radars. The states estimated are the osculating classical orbital elements and the satellite ballistic coefficient. The dynamics used to propagate the state vector forward include the two-body acceleration plus perturbations due to atmospheric drag, zonal harmonics in the geopotential through JS, and tesseral harmonics in the geopotential through J44. The atmospheric density model used is an exponential model that includes diurnal variations and variations in the decimeter solar flux. The observations used to update the state vector estimates are slant range, azimuth, and elevation relative to a radar site. The three filters investigated in this research are a nonlinear least squares filter, an Extended Kalman filter, and a Gauss second order filter. Data are processed for three different satellites. The first is a high altitude (1000km at perigee), non-circular (e=0.015), orbit. The second satellite orbit is a low altitude (250km at perigee), non-circular (e=0.01), orbit. The final orbit is a low altitude (300km), nearly circular (e=0.0003), orbit. The filters are compared using four criteria: estimation errors, prediction errors, computer time of operation, and computer storage requirements. The Gauss second order filter is shown to provide a substantial improvement in orbit determination accuracy for satellites subject to significant perturbing accelerations.

Original languageEnglish (US)
StatePublished - Jan 1 1986
EventAstrodynamics Conference, 1986 - Williamsburg, United States
Duration: Aug 18 1986Aug 20 1986

Other

OtherAstrodynamics Conference, 1986
CountryUnited States
CityWilliamsburg
Period8/18/868/20/86

Fingerprint

nonlinear filters
orbit determination
filters
orbits
geopotential
state vectors
low altitude
tesseral harmonics
satellite tracking
zonal harmonics
atmospheric density
solar flux
satellite orbits
orbital elements
diurnal variations
Kalman filters
high altitude
azimuth
ballistics
drag

ASJC Scopus subject areas

  • Astronomy and Astrophysics

Cite this

Boden, D. G., & Conway, B. A. (1986). A comparison of nonlinear filters for orbit determination. Paper presented at Astrodynamics Conference, 1986, Williamsburg, United States.

A comparison of nonlinear filters for orbit determination. / Boden, Daryl G.; Conway, Bruce A.

1986. Paper presented at Astrodynamics Conference, 1986, Williamsburg, United States.

Research output: Contribution to conferencePaper

Boden, DG & Conway, BA 1986, 'A comparison of nonlinear filters for orbit determination', Paper presented at Astrodynamics Conference, 1986, Williamsburg, United States, 8/18/86 - 8/20/86.
Boden DG, Conway BA. A comparison of nonlinear filters for orbit determination. 1986. Paper presented at Astrodynamics Conference, 1986, Williamsburg, United States.
Boden, Daryl G. ; Conway, Bruce A. / A comparison of nonlinear filters for orbit determination. Paper presented at Astrodynamics Conference, 1986, Williamsburg, United States.
@conference{3966e6f93d2b4eeb84dda63d6fe88e2e,
title = "A comparison of nonlinear filters for orbit determination",
abstract = "This paper compares the performance of three filters which have been applied to the problem of orbit determination using actual satellite tracking data obtained from ground based radars. The states estimated are the osculating classical orbital elements and the satellite ballistic coefficient. The dynamics used to propagate the state vector forward include the two-body acceleration plus perturbations due to atmospheric drag, zonal harmonics in the geopotential through JS, and tesseral harmonics in the geopotential through J44. The atmospheric density model used is an exponential model that includes diurnal variations and variations in the decimeter solar flux. The observations used to update the state vector estimates are slant range, azimuth, and elevation relative to a radar site. The three filters investigated in this research are a nonlinear least squares filter, an Extended Kalman filter, and a Gauss second order filter. Data are processed for three different satellites. The first is a high altitude (1000km at perigee), non-circular (e=0.015), orbit. The second satellite orbit is a low altitude (250km at perigee), non-circular (e=0.01), orbit. The final orbit is a low altitude (300km), nearly circular (e=0.0003), orbit. The filters are compared using four criteria: estimation errors, prediction errors, computer time of operation, and computer storage requirements. The Gauss second order filter is shown to provide a substantial improvement in orbit determination accuracy for satellites subject to significant perturbing accelerations.",
author = "Boden, {Daryl G.} and Conway, {Bruce A.}",
year = "1986",
month = "1",
day = "1",
language = "English (US)",
note = "Astrodynamics Conference, 1986 ; Conference date: 18-08-1986 Through 20-08-1986",

}

TY - CONF

T1 - A comparison of nonlinear filters for orbit determination

AU - Boden, Daryl G.

AU - Conway, Bruce A.

PY - 1986/1/1

Y1 - 1986/1/1

N2 - This paper compares the performance of three filters which have been applied to the problem of orbit determination using actual satellite tracking data obtained from ground based radars. The states estimated are the osculating classical orbital elements and the satellite ballistic coefficient. The dynamics used to propagate the state vector forward include the two-body acceleration plus perturbations due to atmospheric drag, zonal harmonics in the geopotential through JS, and tesseral harmonics in the geopotential through J44. The atmospheric density model used is an exponential model that includes diurnal variations and variations in the decimeter solar flux. The observations used to update the state vector estimates are slant range, azimuth, and elevation relative to a radar site. The three filters investigated in this research are a nonlinear least squares filter, an Extended Kalman filter, and a Gauss second order filter. Data are processed for three different satellites. The first is a high altitude (1000km at perigee), non-circular (e=0.015), orbit. The second satellite orbit is a low altitude (250km at perigee), non-circular (e=0.01), orbit. The final orbit is a low altitude (300km), nearly circular (e=0.0003), orbit. The filters are compared using four criteria: estimation errors, prediction errors, computer time of operation, and computer storage requirements. The Gauss second order filter is shown to provide a substantial improvement in orbit determination accuracy for satellites subject to significant perturbing accelerations.

AB - This paper compares the performance of three filters which have been applied to the problem of orbit determination using actual satellite tracking data obtained from ground based radars. The states estimated are the osculating classical orbital elements and the satellite ballistic coefficient. The dynamics used to propagate the state vector forward include the two-body acceleration plus perturbations due to atmospheric drag, zonal harmonics in the geopotential through JS, and tesseral harmonics in the geopotential through J44. The atmospheric density model used is an exponential model that includes diurnal variations and variations in the decimeter solar flux. The observations used to update the state vector estimates are slant range, azimuth, and elevation relative to a radar site. The three filters investigated in this research are a nonlinear least squares filter, an Extended Kalman filter, and a Gauss second order filter. Data are processed for three different satellites. The first is a high altitude (1000km at perigee), non-circular (e=0.015), orbit. The second satellite orbit is a low altitude (250km at perigee), non-circular (e=0.01), orbit. The final orbit is a low altitude (300km), nearly circular (e=0.0003), orbit. The filters are compared using four criteria: estimation errors, prediction errors, computer time of operation, and computer storage requirements. The Gauss second order filter is shown to provide a substantial improvement in orbit determination accuracy for satellites subject to significant perturbing accelerations.

UR - http://www.scopus.com/inward/record.url?scp=84997712956&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84997712956&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:84997712956

ER -