ASI/EUMETSAT Meeting

ASI Space Geodesy Center — Matera, 04-05 February 2009

S. Casotto, P. Zoccarato, A. Nardo, M. Bardella

CISAS – University of Padua

SWOrDSoftWare for Oceansat2 Orbit Determination

SWOrD: Software purposes

Multi–satellite LEO-MEO orbit determination based on GPS measurements

LEO-MEO orbit prediction

Modeling and prediction of Radio Occultation (RO) events

High precision reconstruction of SST velocities

Adoption of HPC techniquesComputation kernel separated from the process communications layerLow latencyPOD:

DynamicReduced DynamicKinematics

ROSAROSSA(Research and OperationalSatelliteand SoftwareActivities)data flow

SWOrD: Programming paradigmsObject Oriented Programming (OOP)

Singleton pattern to restrict instantiation of a class to one object

C++/Fortran interoperability

Multi threads/OpenMP parallelization

Database on RAM for input data administration based on Boost::Multiindex library

GPStk: fundamental and advanced GPS processing algorithms

XML input and output manipulation with data binding through Code Synthesis utility

Development and debugging: Eclipse 3.1, Intel C++/Fortran compiler 11 and the software construction tool SCons

UTC: Coordinated Universal TimeUT: Universal timeET: Ephemeris Time. Was used 1960-1983TDT: Terrestrial Dynamical Time. Was used 1984-2000TT: Terrestrial TimeTAI: International Atomic Time (Temps Atomique International)GPS time = TAI - 19 seconds

OOP advantages (1/2)

UTC 1972 -

An example:the Time_Tag class

ET 1960-1983 TDT 1984-2000

TT 2001- UT1

delta-UT = UT1-UTC(max 0.9 sec)

delta-T = TT-UT1

GPS 1980 - TAI 1958 -

TAI-UTCLeap seconds

TAI-GPS19 s fixed

TT-TAI32.184s fixed

OOP advantages (2/2)An example of inheritance usage: the Ground Station and Satellite classes

Boost::MultiindexThe Boost Multi-index Containers Library provides a class template which enables the construction of containers maintaining one or more indices with different sorting and access semantics

XML data binding

CodeSynthesis XSD is an open-source, cross-platform W3C XML Schema to C++ data binding compiler

SWOrD

DPC

Input/Support/Output

Gravity fields,Tidal models

EOP parameters

Solar System Ephemerides

IGS Data

ROSA Data

Solar Geomagnetic

Data

DG_kDG_1

ROSA ROSSAArchives

Planetary Ephemerides (JPL)Models of gravitational field and terrestrial/oceans tides (JMG3, EGM96, EG4)Ground station solutions (IGS)EOP e Leap seconds (bulletins B e C from IERS, bulletin A from USNO)Solar and geomagnetic activity indexes (NGDC)IGS productsROSA navigation dataOCEANSAT2 attitudeROSA receiver multipath pattern

Ancillary data required

L1 Input & Output

L2 Input & Output

Logical model

Tracking Data

ExternalData

Interface Class

DRIVER

ROSA RinexData

G/S Rinex Data

Products

Interface Class

Interface Class

Executive System

ApplicationClass

ApplicationClass

ApplicationClass

ApplicationClassApplication

ClassApplication

Class

Two main kinds of operational blocks: interface classes and application classes.

The first called application class triggers the reading interface classes to acquire the input data.

During this process the input data are manipulated to obtain a complete set of information that will be later used by the other application classes.

SWOrD contains a mathematical model of the

world used to process observational data. The functional model of the

system can be decomposed into several components,

which are the counterpart to the physical model to be

developed

DPC_Interface

Data_Conditioning

Observation_Simulator

Spacecraft_PropagatorStation_Propagator

Normal_Equation_Handler Product_Generator

Functional model

Flow chart

Products

DPC-DG

Interface

GPS DataPre-processing

Cartesian State Data

Pre-processing

Orbit Determination & Prediction

GPS & LEO Ephemerides

RO EventsTables

L1, L2SNR

Predicted RO Events

Tables

L1, L2 Phase excess

GPS & LEOPredicted

Ephemerides

Determined orbits Predicted orbits

PPP LEO Orbit

Kinematic POD (1/2)

There exist several kinematic orbit determination techniques based on either undifferenced, doubly-differenced or triply-differenced observables, but also on time-differenced phase observables.

All of the kinematic approaches to GPS-based LEO OD, including the one adopted here, make use of the fundamental contribution of data provided by the IGS. These data include GPS SV’s orbits, clock solutions for both SV’s and ground stations, EOP and tropospheric zenith delay (TZD) solutions. All these parameters are held fixed in the solution process.

The strategy adopted in the SWOrD OD system is based on the use of a combination of pseudorange and phase observables differenced in time. The procedure is illustrated on the left

Kinematic POD (2/2)At each observation epochposition is computed from

pseudorange data using the Bancroft algorithm

The position fixes obtained are used in a dynamic orbit fit based on a minimal set

of LEO parameters

The pseudorange observations are screened

for outliers based on the comparison of the

Bancroft solutions with the corresponding positions on the dynamically fitted orbit

Firsttimeonly

Phase observations are screened for cycle-slip detection by iono-free,

almost iono-free (4,-3)(5,-4), and triple iono-free carrier

phase combinations against counterparts pseudorange

Phase observations are corrected for relativistic

GPS clock correction, GPS and LEO satellite antenna

offset and attitude, and GPS satellite antenna phase

wind-up

A precise point positioning (PPP) estimation algorithm is applied to the iono-free pseudorange observations and the time-differenced (between successive observation epochs) iono-free carrier phases. PPP is applied both forward in time from the first observation epoch as well as backward in time from the last observation epoch. The filter is

applied by properly accounting for the correlations between phases at adjacent epochs and

propagating the associated covariance matrix. The two estimated position fixes are finally optimally

combined to obtain the best smoothed position and clock estimates at each observation epoch.

LEO velocities are computed at each

ephemeris epoch based on the time derivative of the

interpolating polynomial of positions.

THANKS FORYOUR ATTENTION