Choose timezone
Your profile timezone:
1st ACTS Experiment Liaison Group meeting
→
Europe/Zurich
ACTS Experiment Liaison Group meeting
Zoom Meeting ID
63827894330
Host
Paul Gessinger
Alternative host
Andreas Salzburger
Useful links
Join via phone
Zoom URL
1. ACTS status and preview (A. Salzburger)
- Since the NIKHEF workshop (April 2026):
- detray and traccc were merged into the ACTS repository
- Gen3 geometry is close to feature-complete, including translation to detray with material
- the legacy seeding was removed
- GBTS was added as a CPU and GPU seeder
- Infrastructure:
- C++23 is the default and C++20 is still supported
- CMake 3.19
- new optional dependencies: GBL, Millepede-II, Arrow/Parquet
- A list of API-breaking changes (April–October 2026) is attached to the agenda
- CPU and GPU code paths remain largely separate. Proposals:
- a campaign to reuse code between them
- more cross-reviewing
- merging the CPU and GPU development meetings into one meeting on Tuesdays
- New initiatives, with contributors welcome:
- alignment (meeting on Mondays at 11h00; requests for alternating slots go to the conveners)
- secondary vertexing
- cosmics reconstruction
- drift chamber support
- AI coding agents are increasingly used. Whoever submits a PR must vouch for its code
- Discussion:
- Gen3 construction examples for TGeo and DD4hep: the ODD example shows several approaches, including a path based on TGeo only. The MuC detectors in k4ActsTracking have used Gen3 since summer 2026
- Gen1 timeline: ePIC raised its ~6-month backwards-compatibility window. Agreed:
- Plan to deprecate Gen1-only types by Q1/2027 (with deprecation warnings that can be suppressed)
- they stay deprecated for at least ~6 months
- they are only removed after checking with the experiments
2. Experiment round table
- Muon Collider (F. Meloni):
- follows ACTS main through k4ActsTracking and Spack
- Gen3 geometry and material mapping
- 4D tracking everywhere
- GNN-based tracking is being explored
- ACTS sets the numpy, Eigen and sympy caps in the stack (not currently a problem). Release tarballs now include the generated headers
- Developments are evaluated for upstreaming. GNN ONNX parts are being consolidated upstream
- SHiP (O. Lantwin):
- FairShip has moved from a private fork to ACTS Core from conda-forge and matches or beats GenFit
- the new framework (Trout) starts directly on Gen3
- Feedback:
- telescope geometry needs coordinate rotations
- no support for straws or drift circles
- plugins and Python bindings are missing on conda-forge
- ACTS team:
- alternative parametrizations are under consideration, with no timeline yet
- build geometry through the blueprint mechanism and report where it lacks control
- discuss the straw line fitter in a developer meeting
- SHiP offered to help with conda-forge packaging
- ATLAS (N. Calace):
- ACTS is used across the full track reconstruction
- each ACTS PR is tested against Athena
- Gen3 for ITk is being finalised, with Gen1 to be decommissioned by the end of 2026
- GPU tracking with traccc is being evaluated for online and offline use
- The lower barrel efficiency comes from the different seeding strategy and is being worked on
- Lohengrin (C. Breuning):
- moving to k4ActsTracking
- setting up the telescope geometry with the blueprint
- next steps: CKF, then GSF
- a candidate test case for an alternative parametrization
- ALICE 3 (I. Altsybeev):
- Several contributions to ACTS repository from ALICE
- ACTS main plus ALICE 3 plugins, with Gen3 geometry
- 4D seeding, tracking and vertexing
- alignment for out-of-plane distortions (presentation at an upcoming alignment meeting)
- a pilot integration in O2 exists
- Action: set up downstream CI against ACTS PRs once O2 integration matures
- CMS (L. Damenti):
- an ACTS-based propagator in CMSSW (v44.0.1) for the material treatment
- working on the momentum-scale bias and performance
- Heads-up: the JSON material format is changing, with a migration path
- Suggestion: try the SympyStepper (faster, same results)
- E320 (R. Urmanov):
- frozen at v36.1 (Gen2) with local patches
- interested in Gen3 if migration effort is acceptable
- moving to a newer version for future reprocessing to be discussed
- ePIC (W. Deconinck):
- v46.8.1
- Seeding2 TripletSeeder validated bit for bit against the Orthogonal seeder
- switching to Gen3 with v47 (off-axis B0 support)
- Open issues:
- seed time takes precedence in the final track parameters when the hits carry no time
- reproducibility across threads (sorted space points)
- passing covfie fields directly
- LUXE (T. Madlener):
- fully on k4ActsTracking with a planar Gen3 blueprint
- wants straight-line parameters
- offered as the simplest test bed for an alternative parametrization
- sPHENIX (J. Osborn):
- v45, Gen1
- the default navigator drops up to 1/3 of TPC clusters under real misalignment, so the DirectNavigator is used
- Millepede alignment is giving good silicon residuals
- Suggestion: model the TPC with volume material plus the DirectNavigator
- Gen3 testing in a separate build
- STCF (X. Ai):
- v32.0
- losing efficiency for long-lived particles with inside-out CKF
- low-pT resolution is worse than GenFit
- investigating the ML ambiguity solver and GBTS
- Pointer: ATLAS LRT experience
- FCC (J. Beirer):
- ACTS main through Key4hep and k4ActsTracking, with ACTS CI testing CLD and ILD
- AMVF on CLD matches or beats LCFIPlus
- material scans: TGeo-based scan plus the new ddsim recorder
- strong interest in secondary vertexing
- AGORA (T. Lin):
- frozen at v35.2 for the TDR
- plans:
- extrapolation to calorimeter and muon system
- GSF for low-pT electrons
- upgrade ACTS to more recent version
- integration with FCCSW via k4ActsTracking
There are minutes attached to this event.
Show them.