How can you track the historical distance between two satellites with updating elsets?

[davidkurtenb]

I am trying to capture the historical distance between two satellites but having trouble writing the logic to update the observed TLE when an updated TLE is made available. For example if have the 4 TLEs listed below but as the distance iterates through the differnt times how can I update to use the most recent TLE if the there is an new observation :

################# SAT 1
sat1_initial = EarthSatellite(
'1 50322U 21129B 22225.62649072 -.00000331 +00000+0 +00000+0 0 9999',
'2 50322 0.1164 164.1148 0002656 102.9776 51.8190 01.00428106002426'
)
sat1_update=EarthSatellite(
'1 50322U 21129B 22226.61217274 -.00000326 +00000+0 +00000+0 0 99990',
'2 50322 0.1173 162.8746 0002740 104.0896 48.3189 01.00427772002433'
)

################# SAT 2
sat2_initial = EarthSatellite(
'1 52940U 22073A 24074.90684682 -.00000373 +00000+0 +00000+0 0 99995',
'2 52940 1.5815 89.4725 0002301 223.4944 305.7281 01.00274601006105'
)
sat2_update=EarthSatellite(
'1 52940U 22073A 24075.04354795 -.00000372 +00000+0 +00000+0 0 99999',
'2 52940 1.5818 89.4689 0002312 223.3247 355.2499 01.00274549006094'
)

import numpy as np
import matplotlib.pyplot as plt
from skyfield.api import Topos, Loader, EarthSatellite

ts = load.timescale()
de421 = load('de421.bsp')
earth = de421['earth']

datetime_sample = sat1_initial.epoch.utc_datetime()
year = datetime_sample.year
month = datetime_sample.month
day=datetime_sample.day
hour=datetime_sample.hour

hours = np.linspace(hour,hour+24, 25)
times = ts.utc(year, month, day, hours, 0)
times_lst = sorted(times.utc_strftime())
print(times_lst)

separation_lst = []

sat1_pos = sat1_initial.at(times).position.km
sat2_pos = sat2_initial.at(times).position.km
separation = np.sqrt(((sat1_pos-sat2_pos)**2).sum(axis=0))# kilometers
separation_lst.append(separation)
change = separation[1:]-separation[:-1] # kilometers per minute
range_rate = change * 1E+06 / 60 # millilmeters per second

print(len(separation_lst))

[brandon-rhodes]

Since this doesn't involve a bug in Skyfield, I'm going to convert it into a Discussion, so that you can get help with your script in a Q&A format.

[brandon-rhodes]

Let's focus first on refining your question. The script as written already knows the "better", more recent, TLEs, so it would just use them to start with, if I understand your question. But in "real life", where the later = better TLE is not at first available, the new TLE must somehow be delivered or discovered. Could you outline for us how that takes place?

[davidkurtenb]

I'm trying to find the historical distances between two satellites. So for example, if I am looking for the distance between sat1 and sat2 on given date, say 12/21/2019, and provide an EarthSatellite object for a set of TLEs observed on 12/21/19 (the set being one TLE for sat1 and another for sat2). But then want to know the distance on 12/23/2019 (two days later) and there is a newer TLE for sat1, epoch date of 12/22/2019, how would I update the EarthSatellite object to use the new TLE instead of the original all while maintaining the list of distances between the two?

Or am I misunderstanding how the .at function works? For example, if I provide an EarthSatellite object and .at the object will that automatically go find the most recent TLE observation for the given satellite based on the time search?

[brandon-rhodes]

Ah, understood, this is for historical work! You are correctly understanding .at(): it turns a single TLE into positions, without any magic behind to scenes to switch to another TLE.

My guess is that you will want, for each satellite, to sort its TLEs by epoch; then compute the midpoint between each pair of adjacent epochs to decide on a crossover date-and-time where you'll make the switch to the new TLE; then split your time array at those crossover points; then call .at() for each span of times, calling the satellite object whose epoch is at the center of those times; and then put the resulting coordinates back together.

An immediate problem I can think of is of getting a jump in position when you switch, maybe a big one? So you might need overlap time periods where you somehow transition gradually from one set of coordinates to another? But I'm not familiar with how that's done in the industry, so that the intermediate positions are plausible.