add ActsTrkFinding to use ACTS CKF

A first implementation of ActsTrkFinding/TrackFindingAlg and ActsTrkFindingTool/TrackFindingTool to use Acts::CombinatorialKalmanFilter from Athena.

• With --preInclude ActsInterop.ActsCIFlags.actsValidateTracksFlags:
  • replaces Athena track finding with Acts track finding. Subsequent steps use Acts Trk::Track outputs, so default IDPVM can be used for performance studies.
    • It looks good so far, with one big exception: a large loss of efficiency in the barrel, probably due to loss of most strip hits.
  • also replaces the Acts clusterization with the Athena clusterization in the Acts workflow. This is needed to create PRD links, which are needed while we use Trk::Track as output of the track finding.
• To-do following this MR:
  • fix strip hit association
  • revisit tool interface, or move tool functionality into algorithm.
  • improve Trk::Track creation or replace with new Acts::Track. For Trk::Track:
    • does not add outliers from reverse filtering - is this needed?
    • uses legacy Athena enums, Trk::TrackInfo::TrackFitter::KalmanFitter and Trk::TrackInfo::TrackPatternRecoInfo::SiSPSeededFinder
    • only works with Athena clusterization
    • Unlike Athena algorithm, no additional fit required after Acts CKF. We should disable the global Chi^2 fit in the ambiguity solver.
    • are we using the correct TSOS to calculate the track parameters?

Implements ATLIDTRKCP-355. Details and performance discussed here.

assigned to @adye

added 1 commit

• 33ff3fd7 - get geometry and B-field based on ActsKalmanFitter

Compare with previous version

added 1 commit

• e9b45cb3 - Fix interface to ACTS CKF. Now it compiles!, but doesn't run, or I/O seeds/tracks

Compare with previous version

In order to test this implementation within athena, as discussed in our chat (with @pagessin), we could think of making the ACTS CKF pluggable. It requires a tool that public extends<AthAlgTool, ISiTrackMaker> and implements the functions needed in SiSPSeededTrackFinder::newStrategy. This would probably need some gymnastic to convert from the current to the new EDS. Is this something you plan to add in this MR as well?

added 1 commit

• 6af142bf - fix TrackFindingTool's tool setup. Now compiles and runs, 'just' doesn't see input seeds.

Compare with previous version

Is this something you plan to add in this MR as well?

I think it would be good if this MR can be merged with something that works, even if not the final interface. So I would prefer to make the new interface as a separate MR.

However, if the new frozen T0 policy means this can't be merged anyway, then we could continue development in this MR.

Another issue is that I am working with 23.0.9 at the moment. At some point I will have to update to be compatible with Carlo's updates for v21.0.0/v21.1.0. I know there are some changes in the ACTS CKF API because I had to remove those to get it to compile in 23.0.9. There are probably some EDM changes I'll need to update too.

Maybe now Carlo's MRs have been merged, I should update sooner rather than later. It's not in a 23.0.X release yet, so I'll need to use a nightly which isn't as stable.

added 1 commit

• 4fcf2e04 - use estimatedTrackParameters from seeding

Compare with previous version

added 1 commit

• 697d1370 - read pixel/strip clusters and convert measurements to sourcelinks for CKF....

Compare with previous version

added 1 commit

• 5cee845e - remove unused code

Compare with previous version

added ACTS label

Hi @adye from a first look I have a few comments on the code.

Concerning the support for the ATLAS interface, I'd suggest to follow the same approach used for the Seeding (see https://gitlab.cern.ch/atlas/athena/-/blob/master/Tracking/Acts/ActsTrkTools/ActsTrkSeedingTool/src/SiSpacePointsSeedMaker.h)

Thanks for the idea. I'll take another look there.