Orbit Determination using PVT input by SequentialBatchLeastSquares

Hello everyone, I have recently been trying to implement an orbit prediction algorithm using the SequentialBatchLeastSquares example from Orekit. Currently, I have GPS data for the satellite
gps-02280302 (114.5 KB)
, but during the process, the following exception was thrown.There are some outlier points in the data, for example, having the same PV value at different times. I am sure that these points have been filtered out, but the problem still persists. Please help me to confirm the issue. Also, if you have any suggestions on using GPS data for orbit extrapolation, please let me know. Thank you very much.

Exception in thread "main" org.orekit.errors.OrekitException: NaN appears during integration near time NaN
	at org.orekit.errors.OrekitException.unwrap(OrekitException.java:154)
	at org.orekit.propagation.integration.AbstractIntegratedPropagator.integrateDynamics(AbstractIntegratedPropagator.java:511)
	at org.orekit.propagation.integration.AbstractIntegratedPropagator.propagate(AbstractIntegratedPropagator.java:408)
	at org.orekit.propagation.PropagatorsParallelizer.propagate(PropagatorsParallelizer.java:140)
	at org.orekit.estimation.leastsquares.AbstractBatchLSModel.value(AbstractBatchLSModel.java:319)
	at org.hipparchus.optim.nonlinear.vector.leastsquares.LeastSquaresFactory$LocalLeastSquaresProblem.evaluate(LeastSquaresFactory.java:440)
	at org.orekit.estimation.leastsquares.BatchLSEstimator$TappedLSProblem.evaluate(BatchLSEstimator.java:615)
	at org.hipparchus.optim.nonlinear.vector.leastsquares.GaussNewtonOptimizer.optimize(GaussNewtonOptimizer.java:163)
	at org.orekit.estimation.leastsquares.BatchLSEstimator.estimate(BatchLSEstimator.java:435)
	at org.orekit.tutorials.test.SequentialBatchLeastSquaresTest.run(SequentialBatchLeastSquaresTest.java:209)
	at org.orekit.tutorials.test.SequentialBatchLeastSquaresTest.main(SequentialBatchLeastSquaresTest.java:157)
Caused by: org.hipparchus.exception.MathIllegalStateException: NaN appears during integration near time NaN
	at org.hipparchus.ode.nonstiff.EmbeddedRungeKuttaIntegrator.integrate(EmbeddedRungeKuttaIntegrator.java:269)
	at org.orekit.propagation.integration.AbstractIntegratedPropagator.integrateDynamics(AbstractIntegratedPropagator.java:477)

Hi @toonaiveckuu

Welcome to the Orekit forum!

What is the frame of your measurements?

It is important that the frame used to use the same frame between the measurement and the orbit propagator used for estimation.

Also, what is the time scale used for the epochs in the file?

If you can’t solve the issue based on the two previous questions, could you provide a sample of code to check what could be wrong?

Best regards,

1 Like

Thank you for your reply. I found the reason for the exception ‘NaN appears during integration near time NaN.’ It was because I set the weight of the measurement values incorrectly.
However, after running the code, the deviation between the result and the measured values is very large. I used the SequentialBatchLeastSquares example in orekit-tutorials-11.3, and changed the type of Measurements to PV.
frame of measurements is EME2000
The following is the code.

pvData satellite and weights are configured in a YAML file.

  # Measurements definition
    measurementFiles: [ "gps-02280302.txt","gps-0228.txt" ]
      sigmaPos: 10.0
      weight: 1.0
      sigmaPos: 10.0
      sigmaVel: 1.0
      weight: 1.0

This is the result and code

this is the config file

durationInDays: 7.0
outputStep: 300.0

  # Orbit definition (position in meters, velocity in meter per seconds, angles in degrees, date in ISO 8601)
    date: "2023-02-28T12:19:12.000"
    frameName: "EME2000"
      name: "CARTESIAN"
        x: -4519721.5
        y: 3015117.9
        z: 4136884.9
        vx: 4712.32
        vy: -1191.83
        vz: 5998.61
  # Spacecraft definition (mass in kilogrammes)
    name: "QL1"
    id: "1992-070A"
    mass: 192

  # Propagator definition
    # Numerical integrator (fixed step (s) and position error (m))
      minStep: 0.001
      maxStep: 300.0
      positionError: 100.0
    # Force models used by the propagator (only the ones used !!!)
      # Central body gravity
        degree: 20
        order: 20
      # 3rd body attraction
        - name: "Sun"
          withSolidTides: false
        - name: "Moon"
          withSolidTides: false
      # Post-Newtonian correction force due to general relativity ("isUsed: true" to not have an empty value)
        isUsed: true

  # Body definition (default is a WGS-84 ellipsoid with IERS-2010 conventions and simple EOP frame)
    iersConventionYear: 2010
    frameName: "CIO/2010-based ITRF simple EOP"
    equatorialRadius: 6378137.0
    inverseFlattening: 298.257223563

  # Measurements definition
    # BE CAREFUL: With the current implementation of the tutorial, measurement files must be chronologically ordered
    measurementFiles: [ "gps-0228.txt","gps-0228.txt" ]
      sigmaPos: 10.0
      weight: 1.0
      sigmaPos: 10.0
      sigmaVel: 1.0
      weight: 1.0

  # Estimator definition
    orbitalParametersPositionScale: 100.0
    # Levenberg-Marquardt or a Gauss-Newton
      # levenbergMarquardt or gaussNewton
        initialStep: 1.0e6
    maxIterations: 100
    maxEvaluations: 100
    convergenceThreshold: 1.0e-3

Could you try using NumericalOrbitDetermination tutorial?

Before running it, could you:

  • Remove one of the two file in the following line: measurementFiles: [ "gps-0228.txt","gps-0228.txt" ]
  • Reduce the velocity sigma from 1.0 m/s to 0.01 m/s

If you still have issues, next steps are, I think, to investigate the frame and the time scale used in the measurement file.


Thank you for your reply. I will try it out and get back to you.