Closed mattbrown11 closed 1 year ago
The comparison between the current numerically calculated Jacobian and the new analytic Jacobian, assert np.allclose(jac, jac0, atol=5e-6, rtol=1e-5)
passed on my system, but the CI failed, likely due to different numerical errors. So, I reduced to assert np.allclose(jac, jac0, atol=5e-4, rtol=1e-5)
. The current Jacobian isn't a "truth" value, but we'd expect it to be close to the new value.
Implemented an analytic expression for the Jacobian of CartesianToElevationBearingRangeRate. This is a more-accurate calculation than the baseline numerical Jacobian calculation. Also, on my system, this sped up the calculation from 500 usec to 100 usec.