All the versions of this article: English , français
* continue comparing algorithms (similar and different)|
* find & try out new algorithms
* combine good features of existing algorithms
* try out algorithms on simulated AGATA data (with all physical effects & dead materials)
* produce data for data analysis working group
* online implementation
* anything else you can think of.....
In this last point one can put the items which came up during the tracking
talk and during various team meetings and presentations:
* get "realistic" position uncertainties from simulations/scanning of a
real cristal |
* improve figures of merit used in the algorithms (they should have a physical meaning)
* check out what algorithms are developped in other fields (gamma-ray astronomy, medical imaging.....)
* track neutrons
During the team meeting, there was a discussion on the way the data from
simulations is pre-processed. In particular, how the packing and the
position smearing (uncertainty + smearing) are done and what kind of
energy resolution one applies.
|It turns out that the packing done by most groups has to be revised so as not to "not pack" points: If the packing is done on a 2 interation step basis, the centroid of the newly packed points may move too far (> 5mm) away from a point which should also be packed.|
Up to now, the uncertainty in the position of an interaction point is
taken to be equal in the x,y and z directions and is only energy
dependent. This is not entirely correct since from the PSA/detector
characterisation teams, we know that certain positions in the detector
are better know than others.
Concerning the data itself, "new trackers" would like to be able to know
what the interaction mechanism is for each interaction point in order
to test their algorithms and have a better idea of what is going on.
Enrico Farnea said that this could easily be implemented in the output of
his GEANT4 simulation code. This could be an extra column, for example,
with a number indicating photoelectric, compton, pair production and rayleigh.
|We need to agree on a "common" data format so that all tracking programs have the same input format and developpers don’t need to worry about changing their programs to read the data....|
Data files exist already (some raw, i.e not packed or smeared and with no
energy threshold and some already pre-processed).
Raw data for an ideal Ge sphere (15-24 cm inner and outer radius) can be
found on Dino Bazzacco’s web site.
Smeared and packed data exist for the same sphere on Franco Camera’s web
Also, Enrico’s Geant4 code is available for anybody who wishes to produce
their own data with various geometries (sphere, simple Ge, agata 120,
agata 180......). You should contact Enrico for more info or go directly to the website: http://agata.pd.infn.it/agata_simul.htm . Simulated data sets are also availbale on that site.
Since in all these simulations, the electron momentum profile is not included (cf
Larissa Milechina’s talk ), one should include it so as to see what effect
it has on the various algorithms (we know already that it has nearly 0
effect on the forward tracking and this is due to the way the interaction
points are weighted in a track and the Stockholm results show that the backtracking algorithm suffers a lot from it).
|New data files should be produced with the electron momentum profile effect.|
This is easily done with Enrico’s code by running with the -C (for Compton profile) option.
For the full agata simulation, the different possible geometries (120
triple or quadruple cluster, 180 triple cluster, 180 single capsules....)
need to be compared in a "complete" way . It is worthwhile to apply Paola
Spolaore’s Q factor determination (cf Paola’s talk ) to get
something more than just P/T and efficiency. For this, spectra need to
be produced so that Paola Spolaore can apply her method.
At the end of the team meeting, there was a "tour de table" to see who was
doing what or interested in doing what.
|Generally, the single interaction treatment procedure needs improvement. This procedure is COMMON to all algorithms.|
There are 2 people developping fuzzy tracking algorithms (C. Rossi Alvarez, see his talk , and D. Bucurescu). It seems that these algorithms are
successfull in determining the 1st and last position in a track quite
successfully. But the next steps in the tracking procedure need to be improved.
Dino’s forward tracking algorithm (mgt - cf Franco Camera’s talk ) needs to be improved in the sense that
the Chi2 used has no physical meaning and that only a certain type of
pair production events is taken into account (the clusters in which the
2 511 gamma-rays produced by the positron annihilation do not "escape"
from the cluster). P.G. Bizetti proposes to investigate further the pair
There are 2 backtracking programs (one in CSNSM and one in
Stockholm). They need to track the same data in order to be compared.
Results of forward and backtracking (CSNSM) algorithms exist on Franco
Camera’s web site. The way the reconstructed data is produced allows to
compare in great detail the 2 methods (original photon event number, track
order, ....). I.e not only can we compare P/T and efficiency but also
check out whether the algorithms reconstruct different tracks and if so,
which kind of tracks to see if their is a recurrent pattern...
This comparison should be systematically applied to all algorithms so that
we can try and combine the different algorithms into one.
The probabilistic method developped by Strasbourg (cf G. Duchene’s talk ) for the planar
configuration needs to be applied to the AGATA geometry. Enrico’s output
contains information on segments and this can be used to collect
all the energy depositions of one segment into one in order to
apply the method.
Neutron tracking has been looked into by the Uppsala group (see J. Ljungvall’s talk ). The pulse
shapes don’t seem to be too different but maybe there is a way to
discriminate neutrons from gammas in the tracking procedure.
Finally, there was, in the past, an attempt by T. Kroll to use neural
networks to track but this wasn’t followed up. Many people who signed
the attendance sheet seemed interested to look further into this
Also, as O. Stezowski pointed out, it might be worthwhile to go have a
chat with computer science people in your universities/institutes to
submit our problem to them. In the same logic, having a chat with
astronomy/astrophysics/... people will no doubt bring new ideas to
our tracking problem.
The team will meet again at the agata collaboration meeting in march in
Orsay. Although the development of tracking algorithms is not as urgent
as other things in the AGATA project, work NEEDS to be done and we NEED
to increase the efficiency of the tracking.
return to top
MINI AGATA WEEK:21-24/06/2004 in Orsay:
monday: Tracking and simulation of key experiments
tuesday: Ancillary detectors
wednesday: Infrastructure: slow control
thursday: Data Acquisition
PRELIMINARY PROGRAM OF "TRACKING AND SIMULATION OF KEY EXPERIMENTS":
|MONDAY 21/06 MORNING:|
technical key experiment simulation meeting on:
*what kind of input/output format to the tracking programs
*what kind of sorts should be done directly, if any
*who does what ?
|MONDAY 21/06 AFTERNOON:|
tracking meeting in itself on:
*new algorithms - new ideas (if any ?)
*position uncertainty <= could be in output from Agata
code ? or read from lookup table ?
*background reduction with tracking algorithms (suppression
of gamma-rays not comming from the "real" source)
*figures of merit - algorithm optimisation (could be
something to do before every experiment...since event types will be
different from one experiment to the next....)
*pair production event tracking (anything new ?)
*single interaction points (anything better than what we
have now ?)
*ancillary information (could be (re)discussed tuesday)
*online implementation (could be (re)discussed thursday)
*who does what ?
The minutes of the joint simulation of key experiments & tracking meeting are posted here .
AGATA WEEK: 21-25/11/2005 at IReS :
The status of the team work was presented by A. Lopez-Martens.
The minutes of the tracking team meeting are posted here
The presentations by G. Suliman and J. Robin at the tracking team meeting can be found on the program page of the Strasbourg Agata week.
return to top
AGATA WEEK: 6-9/06/2006 at Liverpool :
No minutes were taken at the team meeting. The status of the team work was presented by A. Lopez-Martens. There were then 3 presentations, the first by S. Tachenov on tracking for DESPEC , the second by E. Legay to explain the integration of codes and routines into Narval , the third one by O. Stezowski about the simulation chain .
return to top
AGATA WEEK: 15-19/01/2007 at Orsay :
No minutes were taken at the team meeting. There were 3 presentations: tracking of pair-production events, new clusterisation method and status on tracking algorithms for DESPEC. All these talks and the summary talk can be found on the Agata week program (monday afternoon for the team meeting and tuesday afternoon for the plenary session).
return to top
AGATA WEEK: 12-16/11/2007 at Legnaro :
No minutes were taken. There were several presentations: we had an update on the new clusterisation technique developped in Strasbourg by F. Didierjean and on the tracking algorithm developped for DESPEC by S. Tachenov. A new method using tracking and imaging techniques to characterize the performance of PSA was presented by G. Suliman. The slides of the tracking session can be found here .
return to top
AGATA WEEK: 8-11/08/2008 in Uppsala :
No minutes were taken. There were 2 presentations given at the tracking session. The slides of the talks can be found on the Agata week site on
the tuesday afternoon session .