Main differences between V4.0 and V3.4.1

De Wiki
Aller à : navigation, rechercher

PATRIUS V4.0 is a major release and has a simplified architecture: its former dependencies (Commons-Math, OREKIT and the addons) have been merged into a single standalone library (patrius_V4.0.jar). Users only need to download one library for compilation purpose. To help migrate sources based on PATRIUS 3.x, a script is available (search & replace all the import) and make it quite easy to update to PATRIUS 4.0. The script will transform all Java files found under the current directory.

Java 7 compatibility

PATRIUS is now a Java 7 project : all users need a JRE (or JDK 7) (or higher).

Constant/continuous maneuvers

ConstantThrustManeuver & VariableThrustManeuver have been merged in ContinuousThrustManeuver. Constructors use PropulsiveProperty object to set all parameters. Besides, the behavior and the numerical results are the same as before.

Bug fixes

  • AtmosphereData uses molar mass in kg/mol(and not in g/mol). GlobalAeroModel results are affected.
  • Assembly can now correctly handle mobile parts
  • [STELA] propagator can use osculating parameters for its initial orbit.
  • Some optimizations in:
    • AbsoluteDate: around 5% gain in a lot of algorithms
    • HelmholtzPolynomial
    • GTOD frame : a big computation time issue has been fixed. All the frames of this branch now have the same performances as the IERS one.
  • Normalized Earth potential coefficients can be used for order&degrees above 86 (numerical quality bug fix)

API incompatibilities

Regressions

  • GlobalAeroModel: some regression (about 10 cm after a 6h LEO propagation)
  • Use of MultiNumericalPropagator & Assembly: small regression have been observed (~ 1e-14m)
  • CnCookModel and CtCookModel: theses models have been fixed (former results were wrong)
  • ContinuousThrustManeuver: some numerical regression (ie : 1e-7m in position) can be observed while propagating through a maneuver
  • Earth potential : for high order/degree, there are some numerical regression (1e-7m after 24h numerical propagation of LEO orbit) but a better precision (see § bug fixes).

Return to the introduction ↑