Fast and Reliable Computation of Mean Orbital Elements for ...
Transcript of Fast and Reliable Computation of Mean Orbital Elements for ...
Fast and Reliable Computation of Mean Orbital Elements for Autonomous Orbit Control
Emmanuel Gomez (1), Pablo Servidia (2)* and Martin Espaรฑa (2)**
2nd IAA Latin American Symposium on Small SatellitesNovember 2019, 11th to 16th
(1) Unidad de Formaciรณn Superior โ CONAE, Ruta C45 Km 8, (5187) Provincia de Cordoba, Argentina
(2) Comisiรณn Nacional de Actividades Espaciales, Av. Paseo Colon 751, (1063) CABA, Argentina.
(*) [email protected] (**) [email protected]
Table of contents
โข Why Autonomous Orbit Control?
โข Introduction
โข The Osculating to Mean Problem
โข Proposed Algorithm
โข Preliminary Computations and Definitions
โข Estimation of Mean Orbit Elements with Eckstein-Ustinov model
โข Refinement on the Mean Semi-Major Axis
โข Summary of the Proposed Algorithm
โข Applications
โข Application to Satellite Orbit Determination
โข Application to Relative Orbital Elements Determination
โข Application to Autonomous Orbit Control for Formation Flying
โข Conclusions
Why Autonomous Orbit Control?
โข Minimizes ground orbital maintenance
operations.
โข Uniformly close to nominal flight conditions.
โข Maneuvers can optimized as they do not
need to be made over a ground station.
Artistic Representation: TerraSAR-X and TanDEM-X flying in close formation
Orbital Corrections
Usual Approach: Feedback uses
Ground Segment
Near Future:
Feedback in Flight Segment
Autonomous Orbit Control:
From GNSS:
t, r(t),v(t)
r(t),v(t) -> OOE
OOE -> MOE
MOEs -> ROE
Feedback=f(ROE)
Introduction
Some examples of Autonomous Orbit Control:โข PRISMA: Fly formation of two satelliteโข TerraSAR-X and TanDEM-X: Fly formation of two satellite
OOE: Osculating Orbital Element, MOE: Mean Orbital Element, ROE: Relative Orbital Element
The Osculating to Mean Problem
We are looking for a fast and reliable algorithm to solve the following:
Objective: Given the osculating parameters obtained at a time ๐ก from observations (๐, ๐ฃ)
using a Kepler transformation ((๐, ๐ฃ) โ (๐, ๐, ๐, ฮฉ, ๐, ๐)) we have:
obtain the corresponding (short-period) mean values:
without using past or future data.
๐ ๐๐ ๐: = ๐๐๐ ๐, ๐๐๐ ๐, ๐๐๐ ๐ , ฮฉ๐๐ ๐ , ๐๐๐ ๐, ๐๐๐ ๐๐
๐ := ๐, ๐, ๐, ฮฉ, ๐,๐๐
GNSSReceiver & Navigation
GNSS t, r(t),v(t) Here we have the osculating position and velocity vector
Proposed Algorithm
Assumptions on the Userโs Needs:
โข LEO, Cuasi-circular and near-polar orbitsโข Small satellite missionโข Accurancy in semi-major axis ~ 100 [m]โข Minimum computational effortโข Memory-less solution
Proposed Solution:
โข Non iterativeโข No averaging involvedโข Suitable for on-board computing on a small satellite
Simpler Algorithms imply a lower precision on the
mean parameter estimation, but easier to
implement and verify
Preliminary Computations and Definitionsโข Computation of the Osculating Orbital Elements from r(t) and v(t):
Let ๐ป = ๐ ร ๐ฃ be the angular momentum vector and
๐ = ๐ด๐๐ด๐2 โฅ ๐ป โฅ ๐๐๐ฃ, โฅ ๐ป โฅ2 โ๐ โฅ ๐ โฅ
๐ข = ๐ด๐๐ด๐2 (๐3sin(๐) + cos(๐)(๐2cos(ฮฉ) โ ๐1sin(ฮฉ)), )๐1cos(ฮฉ) + ๐2sin(ฮฉ)
๐ธ = ATAN2๐๐๐ฃ
๐๐, 1 โ
โฅ ๐ โฅ
๐,
๐ = ๐ธ โ ๐sin(๐ธ)
๐ = ๐ข โ ๐
โข Computation of the (quasi-nonsingular) Osculating Orbital Elements (OOE):
The Ustinov parameters โ, ๐ and ๐ are defined as follows
โ โถ= ๐ ๐ ๐๐(๐),๐ โถ= ๐ cos(๐),๐ โถ= ๐ +๐
where ๐ = ๐ +๐ is the mean argument of latitud. Note: For near-circular orbits ๐ and M would be undefined, but h, l and ๐ can be used anyway!
The Mean Orbital Elements are obtained as a function of Eckstein-Ustinov disturbances model for each parameter (๐ฟ๐ , ๐ฟโ, ๐ฟ๐, ๐ฟ๐ , ๐ฟฮฉ, ๐ฟ๐):
๐๐๐ธ๐: = ๐0 โ ๐ฟ๐๐๐๐ธ๐: = ๐0 โ ๐ฟ๐ฮฉ๐๐ธ๐: = ฮฉ0 โ ๐ฟฮฉโ๐๐ธ๐: = โ0 โ ๐ฟโ๐๐๐ธ๐: = ๐0 โ ๐ฟ๐๐๐๐ธ๐: = ๐0 โ ๐ฟ๐
Finally the mean Kepler parameters ๐ and ๐ can be also computed as:
๐๐๐ธ๐: = ๐๐๐ธ๐2 + โ๐๐ธ๐
2
๐๐๐ธ๐: = ๐ด๐๐ด๐2 (โ๐๐ธ๐, ๐๐๐ธ๐)
๐๐๐ธ๐: = ๐๐๐ธ๐ โ ๐๐๐ธ๐
Estimation of Mean Orbit Elements with Eckstein-Ustinov model
For the ๐นโs we use the following modified expressions of those proposed by Spiridonova, S., 2012:
๐ฟ๐ = โ3
4๐โ๐บ2๐ฝ0๐0แพโ๐0cos(๐0) + โ0sin(๐0)
แ+cos(2๐0) +7
3๐0cos(3๐0) + โ0sin(3๐0)
๐ฟฮฉ =3
4๐โ๐บ2๐0แพ7๐0sin(๐0) + 5โ0cos(๐0)
แโsin(2๐0) + โ7
3๐0sin(3๐0) + โ0cos(3๐0)
with the definitions:
๐บ2: = โ๐ฝ2๐ ๐2
๐02 , ๐ฝ0 = sin(๐0),
๐โ = 1 โ3
2๐บ2(3 โ 4๐ฝ0)
where ๐ ๐ is the Earth equatorial radius, and:
๐0: = 1 โ ๐ฝ02๐ ๐๐๐ ๐0 โ
๐
2= cos(๐0)
The definitions of ๐ฟ๐, ๐ฟโ, ๐ฟ๐ and ๐ฟ๐ are as in Spiridonova, S.,2012.
Refinement on the Mean Semi-Major Axis (based on Brouwer-Lyddane theory)
๐๐๐ธ๐๐ต๐ฟ: =
๐๐๐ ๐ + ๐พ2๐๐๐ ๐ (3cos2(๐๐๐ ๐) โ 1)
๐๐๐ ๐
โฅ๐โฅ
3
โ1
๐3
+3(1 โ cos2(๐๐๐ ๐))๐๐๐ ๐
โฅ๐โฅ
3
cos(2๐ข๐๐ ๐) )
where:
๐พ2 โ โ๐ฝ22
๐ ๐ธ๐๐๐ ๐
2
, ๐ โ 1 โ ๐๐๐ธ๐2
Notice that we have used ๐๐๐ธ๐ instead of ๐๐๐ ๐ because the osculating eccentricity of disturbed circular orbits becomes very unaccurate as estimator of its mean value.
The mean parameters are computed though the following steps as in Servidia, P., 2019:
โ Inputs: position ๐ and velocity ๐ฃ in ECI frame.
1. Compute the Kepler and Ustinov osculating parameters.
2. Compute the disturbances (๐ฟ๐, ๐ฟ๐, ๐ฟฮฉ, ๐ฟโ, ๐ฟ๐, ๐ฟ๐).
3. Compute the mean orbit elements .
4. Compute the alternative semi-major axis.
โ Result: ๐๐๐ธ๐๐ต๐ฟ, ๐๐๐ธ๐, ๐๐๐ธ๐, ฮฉ๐๐ธ๐, ๐๐๐ธ๐, ๐๐๐ธ๐.
Summary of the Proposed Algorithm
Application to Satellite Orbit Determination
Figure 1: Osculating parameters h versus l(black) and estimates of Eckstein-Ustinovmean parameters (red)
Figure 2: Osculating argument of perigee (black) and Eckstein-Ustinov mean (red) vs. Number of Orbits
HPOP simulation with 40x40 gravity model and
full environment
าง๐: eccentricity vector
Application to Satellite Orbit Determination
Figure 3: Osculating eccentricity (black) and Eckstein-Ustinov mean (red) vs. Number of Orbits
Figure 4: Osculating RAAN (black) and associated mean estimate through Eckstein-Ustinov (red) vs. Number of Orbits
HPOP simulation with 40x40 gravity model and
full environment
Application to Satellite Orbit Determination
Figure 5: Osculating inclination (black) and associated mean Eckstein-Ustinov estimates (red)
vs. Number of Orbits
Figure 6: Osculating semimajor axis (black) and mean estimate obtained from by Eckstein-Ustinov
(red) and finally the corrected value given by Brouwer-Lyddane (green) vs. Number of Orbits
HPOP simulation with 40x40 gravity model and
full environment
The Relative Orbital Elements between a Chief ๐๐ = (๐๐, โ
๐, ๐๐, ๐๐, ฮฉ
๐, ๐
๐) and Deputy ๐๐ =
(๐๐, โ
๐, ๐๐, ๐๐, ฮฉ
๐, ๐
๐) spacecraft (see [5], [6], [7]) can be computed as follows:
๐ฟ๐๐ ๐ โถ= (๐
๐โ ๐
๐)/๐
๐
๐ฟ๐๐ ๐ โถ= ๐
๐โ ๐
๐+ ฮฉ
๐โ ฮฉ
๐cos(๐
๐)
๐ฟ๐๐ ๐ โถ= ๐
๐โ ๐
๐
๐ฟ๐๐ โ โถ= โ
๐โ โ
๐
๐ฟ๐๐ ๐๐ฅ โถ= ๐
๐โ ๐
๐
๐ฟ๐๐ ๐๐ฆ โถ= ฮฉ
๐โ ฮฉ
๐sin(๐
๐)
where (๐๐ฅ, ๐๐ฆ) is a relative inclination vector to describe the relative motion perpendicular to
the orbital plane, while ๐ฟ๐๐(๐) denotes the relative mean longitude between spacecraft
(D'Amico S., 2010).
Application to Relative Orbital Elements Determination
Application to Relative Orbital Elements Determination (Cont)
Example:โข Envisat and a
deputy 900 [m] forward along track
โข Both sharing the same frozen orbit.
Figure 7: ROE between Chief Orbit (precise Envisat data) and a Deputy
900 meters forward, maintaining SSO frozen conditions.
Application to Autonomous Orbit Control for Formation Flying
๐ฟ๐๐ แถ๐ผ = ๐(๐๐ , ๐๐) + ๐ต๐ แถ๐ฃ๐๐๐ก
๐ต๐: =1
๐๐
0 2 0โ2 0 0
)sin(๐๐ )2cos(๐๐ 0)โcos(๐๐ )2sin(๐๐ 0
0 0 )cos(๐๐0 0 )sin(๐๐
แแถ๐ฃ๐๐๐ก = ๐ ๐๐ก ๐,๐(โ๐ต๐+(๐๐ผ๐ฟ๐
๐๐ผ + ๐(๐๐ , ๐๐))
๐(๐๐ , ๐๐) Secular drift assumed in Col(๐ต๐)
Proposed feedback:
ROE evolution model: secular drift plus RTN input control channels แถ๐ฃ๐๐๐ก :
(๐ต๐: RTN control matrix)
๐๐ โ ๐๐ + ๐๐ Mean argument of latitude of the deputy.
Application to Autonomous Formation Flying (Cont.)
Figure 8: Effect of ROE feedback on the relative
longitude ๐ฟ๐๐(๐).
Simulations made with 40x40 Earth gravity model, and full environmental disturbances.
Figure 9: Effect of ROE feedback on the perigee argument.
Conclusions
โข Autonomous Orbit Control may solve usual orbit acquisition and maitenancemaneuvers.
โข We have shown a simple algorithm to find relative orbital elements using thetypical information available onboard with a GNSS receiver on eachspacecraft
โข Simulations show the results for the evaluation of injection accurancy and foran specific Earth Observation satellite.
Contact
Mgter. Emmanuel Walter Gomez
Unidad de Formaciรณn Superior โ CONAEFalda del Caรฑete โ Cรณrdoba โ Argentina
+54-03547-400000 int 1721
ufs.conae.gov.ar
Application to Upper Stage Injection and EvaluationSimulations made here up to J4
Figure 9: Osculating parameters h versus l (black) and estimates of Eckstein-Ustinov mean parameters (red), whose deviations are not signicative afterinjection.
Figure 10: Osculating argument of perigee (black) and Eckstein-Ustinov mean (red), where it can be seen that once the injection is reached there are no signicativedeviations.
Application to Upper Stage Injection and Evaluation (cont.)Simulations made here up to J4
Figure 11: Osculating eccentricity (black) and Eckstein-Ustinov mean (red), where it can be seen that once the injection is reached there are no signicative deviations.
Figure 12: Osculating RAAN (black) and associated mean estimate through Eckstein-Ustinov (red) vs. time in seconds since launch, where it can be seen that once the injection is reached there are nosignificative deviations. Also, the BL mean value (green) is almost equal to the osculating value (black).
Application to Upper Stage Injection and Evaluation (cont.)Simulations made here up to J4
Figure 13: Osculating inclination (black) and associated mean Eckstein-Ustinov estimates (red) vs. time in seconds since launch, where it canbe seen that once the injection is reached there are no significative deviations. Also, the BL mean (magenta) is worse than the EU mean.
Figure 14: Osculating semimajor axis (black) and mean estimate obtained from by Eckstein-Ustinov (red) and finally the corrected value given by Brouwer-Lyddane(green). An additional correction is given in dotted green trace as a result of the adjustment.