Was wondering, in measurement generation, the full package is always computed (like in the orbit determination): position, velocity and acceleration vectors of all participants where the signal rebounds. If I’m not wrong, it’s mostly useful for subsequent corrections (in Orekit terms: modifiers). But it can actually become the bottleneck of the theoretical evaluation, especially with ground based sensors for which it is computationally expensive to retrieve full kinematics states. In some applications, these are completely useless. If you want to generate millions of measurements (@luc) and don’t care about corrections (probably not Luc) you’re wasting a lot of time. So my proposal would be to make this step optional for generators. Thoughts?
The performance for measurements generation is exactly the reason why I set up a few months ago two different methods : theoreticalEvaluation and theoreticalEvaluationWithoutDerivatives in AbstractMeasurement.
The first one is used in orbit determination, whereas the second one is used in measurements generation.
Yeah I know and actually some estimators like the UKF also use the derivative free evaluation. What I’m talking about is the cost of computing fill kinematics states for signal receiver, transmitter, …, emitter. For GroundStation, there is a bottleneck in getOffsetToInertial which builds a Transform. If there was an option to focus on the measurement value which only requires the receiver position for “downlink”, we could use a cheaper StaticTransform.