Parallel Propagation of satellite constellation

Orekit Python Wrapper usage
1

Hi Everyone,

I am writing a Python code which propagates a constellation of 4 satellites for a duration of 7 days using TLE propagator and saves the sub-satellite points of each satellite at an interval of 1 seconds. My goal is to propagate multiple satellites parallelly in a way which takes the least amount of time. Hence, I have considered using multi-threading via PropagatorParallelizer to propagate each satellite independently.

Below is the code I’ve written so far :


### STEP 1 - Setup Parallel Propagation ###

# Initialize arrays
 ...

total_sats = 4
selected_tles = get_tle_data(total_sats) 			#Function to get TLE data of all satellites

# For each satellite, setup propagator 
for satellite_no in range(0, total_sats):

    tle_data = selected_tles[satellite_no] # Get TLE Data for current satellite 
    tle_line1 = tle_data[0]
    tle_line2 = tle_data[1]
	
    inertialFrame = FramesFactory.getTEME()
    mytle = TLE(tle_line1, tle_line2)
    propagator = TLEPropagator.selectExtrapolator(mytle)
	
    propagator_list.append(propagator)
    inertialFrame_list.append(inertialFrame)
    

propagators = Arrays.asList(propagator_list)

class CustomStepHandler(PythonMultiSatStepHandler):

    def init(self, states0, t):
        pass

    def handleStep(self, currentState):
        pass

    def finish(self, s):
        pass

my_handler = CustomStepHandler()
parallelizer = PropagatorsParallelizer(propagators, my_handler)


### STEP 3 - Propagate Satellites ###

t = [initialDate.shiftedBy(float(dt)) \
        for dt in np.arange(0, duration+step_time, step_time)]      # Datetime array - [ start date , end date , step = interval ]
 
print(f"Propagating {total_sats} satellites...")

states = [ parallelizer.propagate(initialDate, tt) for tt in t ]    # Calculate states after each 1 second interval  ( **Bottle-neck step ) 

plist = [[xstate.getPVCoordinates().getPosition() for xstate in state] for state in states] 

#Function to extract LLA from position and write all SSP to individual csv file for each satellite
...

I have a couple of questions :

  1. The current parallel implementation is taking more time ( approximately 8 mins ) than propagating each satellite sequentially in a loop ( which takes 2 mins in total ) .I wanted to understand why that might be happening.

  2. Is there a way to synchronize the individual propagators with the same propagation timestep to reduce the wait time taken by the global handler at the end of each step?

  3. Is there a way to improve the current code to reduce the computation time ? Or if anyone can suggest any alternate method, it would be very helpful.

For clarity, each satellite is propagated for a duration of 7 days with same start date and different initial TLEs.

Thanks in advance!

2

I would strongly suggest to avoid the loop calling propagate(initialDate, tt) with several values of tt and replace it by a proper implementation of step handler. You could do just one propagation and have the handler store the results on the fly.
As written, the code is not really using multi-threading, it just adds the overhead.
In fact, using a step handler is even much more efficient than explicit loops even for single satellite propagation, and I guess this is even more pronounced in multi-sat propagation.