A first implementation is available on the branch tickets/FIBERALLOC-38. It successfully eliminates the trajectory collisions in demo_netflow.py.

mxhf , it would be great if you could have a look!

At the moment, I'm re-running the entire assignment if trajectory collisions are detected. Shortcuts may be possible, but have the potential to introduce further complications. My hope is that, at least once we switch to radial moves, trajectory collisions will be extremely rare, so that we might be able to afford to do the full recomputation.

just checked it out ....

Works well. Though with the new XML I do get an error in the simulator. 

Optimal solution found (tolerance 1.00e-04)

Best objective 8.190310002886e+11, best bound 8.190310002886e+11, gap 0.0000%

Checking for trajectory collisions

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:186: RuntimeWarning: invalid value encountered in true_divide

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:186: RuntimeWarning: invalid value encountered in arccos

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:314: RuntimeWarning: divide by zero encountered in true_divide

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:314: RuntimeWarning: invalid value encountered in true_divide

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:314: RuntimeWarning: invalid value encountered in arccos

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:315: RuntimeWarning: invalid value encountered in true_divide

/Users/mxhf/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CobraGroup.py:315: RuntimeWarning: invalid value encountered in arccos

---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

~/ownCloudRZG/work/MPE/pfs/src/ets_fiberalloc/demo_netflow_ASIAA.py in <module>

    139

    140         simulator = CollisionSimulator(bench, TargetGroup(selectedTargets, ids))

--> 141         simulator.run()

    142         if np.any(simulator.endPointCollisions):

    143             print("ERROR: detected end point collision, which should be impossible")

~/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CollisionSimulator.py in run(self, solveCollisions)

     96

     97         # Calculate the cobra trajectories

---> 98         self.calculateTrajectories()

     99

    100         # Detect cobra collisions during the trajectory

~/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/CollisionSimulator.py in calculateTrajectories(self)

    267                                             finalFiberPositions=self.finalFiberPositions,

    268                                             movementDirections=self.movementDirections,

--> 269                                             movementStrategies=self.movementStrategies)

    270

    271

~/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/TrajectoryGroup.py in _init_(self, nSteps, stepWidth, bench, finalFiberPositions, movementDirections, movementStrategies)

     72

     73         # Calculate the cobra trajectories

---> 74         self.calculateCobraTrajectories()

     75

     76

~/.pyenv/versions/anaconda3-5.0.1/lib/python3.6/site-packages/ics.cobraOps-0.0.0-py3.6.egg/ics/cobraOps/TrajectoryGroup.py in calculateCobraTrajectories(self)

    138             initOffset = np.pi + startPhi[c] if posPhiMovement[c] else np.abs(startPhi[c])

    139             stepLimits = np.interp([initOffset, initOffset + np.abs(deltaPhi[c])], phiOffsets, phiSteps)

--> 140             stepMoves = np.concatenate((np.arange(stepLimits[0], stepLimits[1], self.stepWidth), [stepLimits[1]]))

    141             phiMoves = np.interp(stepMoves, phiSteps, phiOffsets)

    142             phiMoves = phiMoves - np.pi if posPhiMovement[c] else -phiMoves

ValueError: arange: cannot compute length

Hi, I updated the cobraOps master branch and now it should work with the new XML files.

Use this demo as an example:

https://github.com/Subaru-PFS/ics_cobraOps/blob/master/python/ics/demos/demo2.py

Note that now some cobras have problems, so they are not moved and remain in the home position in simulation. You can use bench.cobras.hasProblem to get the cobras with problems.

Another thing is that I constrain the cobras phi angle to be between -pi and 0. Otherwise the (x,y) --> (tht, phi) calculation is degenerate and it requires a more clever way to calculate it. cobraCharmer can handle these situations, but I didn't implement it yet in the simulator code, so I force the phi angle to be between -pi and 0. 

I hope it now works!

I forgot to add that some cobras have very long link lengths. I guess it's a measure problem, but it creates a lot of trajectory collisions.

And finally, I didn't implement yet the radial moves, so the simulator will predict more collisions than with the radial move.

Thank you very much for fixing this Javier & Martin! 

@Javier: demo2 works on my machine, only I noticed a new dependency on ics_cobraCharmer.

For this ticket:

I just committed to FIBERALLOC-38:

• Added netflow_asiaa.py, which is slightly modified version of demo_netflow.py for the PFI testing use case.
• Added target_generator.py which currently implements a homogenuous target distribution only.
• Modified netflow.py to also accept target tables in the astropy.table ecsv format.

Please check the attached allocations file as first version for ASIAA. It was generated with the following commands.

> python target_generator.py --plot -o data/test_sci.dat -r 100000
> python target_generator.py --plot -o data/test_sky.dat -r 1000
> python target_generator.py --plot -o data/test_cal.dat -r 40
> python netflow_asiaa.py

mv output.txt pfi_alloc_test_20201227.txt

pfi_alloc_test_20201227.txt

The most recent netflow_asiaa.py now also writes a PfsDesign fits file.

This is an example (pfsdesign_test_20201228.fits) with a matching ascii file (pfi_alloc_test_20201228.txt).

Max

This images corroborates the issue in the most recent 

2020-10-15-theta-slow.xml

I think.

Max

mxhf I'm currently working at turning the PFSDesign I/O into library functions. So far I'm making good progress; output is more or less done (thanks to your work in demo_asiaa.py!), but for the input I'm wondering where we are supposed to get the exposure times from... The pfsDesign file describes telescope pointing and orientation, but not the observation duration. Any ideas?

Recently this ticked was discussed at the informal meeting between ASIAA, Hassan and PO. By the time of November run, Martin and Max implemented a functionality to write pfsDesign. Some of pfsDesign files were made using (the most) latest version of pfs_utils after the Nov run (and after fixing a few Bug.)

/work/pfs/ets/kiyoyabe/test_2022jan/fiber_allocation/commissioning/design

0x65d65dc76164ff80  --> agcc_20211120_0424219 
0x0514a60eadf01a87  --> agcc_20211120_2042567 
0x47dcd57b6af5aa53  --> agcc_20211120_2057506
0x5d2e53852a8d7d35  --> agcc_20211120_2254174
0x691e373818a88a23  --> agcc_20211120_2303420
0x0513dafb2d419f9f  --> agcc_20211121_1831017
0x316c25cd18ca62b2  --> agcc_20211121_2331497 

chyan could you try either 0x5d2e53852a8d7d35 or 0x316c25cd18ca62b2? They are several science targets. Most of the cobras are not assigned, but I hope it is OK to test.. (If not, let's have one.)

This will be closed based on testing done during the May Eng Run.

The design file generated and populated in system is always checked by ETS. Therefore I would close this ticket. If there is any other problem, we should open another ticket to address the question.