RAVE

1
RAVE a detector-independent
vertex reconstruction toolkit
W. Waltenberger, F. Moser, W. Mitaroff
Austrian Academy of Sciences
Institute of High Energy Physics, Vienna
A detector-independent toolkit for vertex
reconstruction (RAVE) is being developed, along
with a standalone framework (VERTIGO) for
testing, analyzing and debugging. Main design
goals are ease of use, flexibility for embedding
into existing software, extensibility, and
openness. A beta release is available.

• Motivation and goals
• Track and vertex reconstruction must not
  compromise the high spatial resolution of modern
  semiconductor detectors.
• This goal can be achieved by new, sophisticated
  methods beyond the traditional least squares or
  Kalman filter estimators, using robust,
  non-linear adaptive algorithms.
• The early stages of data analysis local pattern
  recognition, track search and track fitting are
  highly detector-dependent, and their software is
  usually coded from scratch.
• The next stage vertex reconstruction is,
  however, almost fully detector-independent, and
  re-coding may be avoided if there exists an
  adequate, reliable and easy-to-use software
  toolkit.
• Such a toolkit should collect the world-best
  algorithms available for vertex reconstruction,
  i.e. finding, fitting and kinematics.
• Its implementation should make extensive use of
  open standards.
• Acronyms
• RAVE reconstruction (of vertices) in
  abstract versatile environments,
• VERTIGO vertex reconstruction toolkit and
  interfaces to generic objects.

• The RAVE / VERTIGO project
• 1. Development of an extensible,
  detector-independent toolkit (RAVE) for vertex
  reconstruction, to be embedded into various
  environments
• RAVE includes the core algorithms for both vertex
  finding (a pattern recognition task a.k.a. track
  bundling) and vertex fitting (estimation of the
  vertex parameters and covariance matrix).
• Starting point was the CMS offline reconstruction
  software (ORCA), coded in C, which has
  recently been refactored and ported to a new
  framework (CMSSW).
• Principal assets are robust reconstruction
  algorithms with estimators based on adaptive
  filters, thus downweighting the influence of
  outliers (tracks not belonging to the vertex
  being fitted).
• Thanks to its generic API, the toolkit may easily
  be embedded into the software environments of
  various experiments (e.g. CMS at LHC, LDC and SiD
  at ILC, BELLE at KEKB).
• Development of a simple standalone framework
  (VERTIGO) for fast testing, analyzing and
  debugging of the core algorithms
• Framework tools available visualisation,
  histogramming, vertex gun for artificial
  events, LCIO input interface, and data
  harvester (abstract persistency solution) for
  flexible I/O.
• Emulation of various detector setups is supported
  by the skin concept, introducing an
  experiment-specific intermediate layer between
  the RAVE core and VERTIGO.

• Outlook
• Near future
• Interfacing RAVE with the ZvTop (topological
  vertex search) algorithm developed for SLD at
  SLC, at present re-coded in C (by RAL) and Java
  (by SLAC)
• Embedding RAVE into the new Marlin version
  (standard L3 track representation)
• Refinement of the VERTIGO skins for the LDC and
  SiD detectors at ILC
• Embedding RAVE into the org.lcsim (Java based)
  ILC reconstruction software by means of a C
  wrapper (SWIG).
• Mid-term future
• Augmenting RAVE with the CMS kinematics fitting
  and b-tagging algorithms
• A simple VERTIGO skin for the BELLE detector
  (Panther track representation)
• Embedding RAVE into the BASF (C based) BELLE
  reconstruction software.
• Long-term future
• Re-coding of RAVE in Java 5, if the C wrapper
  solution proves inefficient.
• Maintenance and development
• We are committed to maintenance, documentation
  and distribution of RAVE and VERTIGO. A WebSVN
  repository exists, containing our beta release.

VERTIGO event loop diagram
RAVE class diagram (vertex factory)
VERTIGO functionality
VERTIGO class diagram (event generator)
VERTIGO class diagram (observer)
Embedding RAVE into non-CMS software
VERTIGO emulating detector setups

• Status
• Full CMS skin (native, i.e. no parameter
  conversions needed) tested, served as a test
  bed as long as CMSSW was not fully operational
• Simple LDC and SiD skins (no materials) tested,
  with input from LCIO (switch for standard L3
  or temporary Brahms track representations).

• Status
• Fully tested within the native new CMS
  software framework (CMSSW)
• Finished for C environments tested with
  MarlinReco/Marlin (temporary solution for
  Brahms track fit parameters and covariance
  matrix)
• Java 5 environments need C wrapper (SWIG)
  first tests with org.lcsim.

Visualisation of adaptively fitted c-cbar and
b-bbar events (simulated for CMS)