Next: Observation Interval Determination for the Chandra X-ray Observatory
Up: Data Pipelines and Quality Control
Previous: Generating Calibration Reference Files with an OPUS Pipeline
Table of Contents -
Subject Index -
Author Index -
PS reprint -
Bringer, M., Boër, M., & Morand, F. 2000, in ASP Conf. Ser., Vol. 216, Astronomical Data
Analysis Software and Systems IX, eds. N. Manset, C. Veillet, D. Crabtree (San Francisco: ASP), 445
A Pipeline for an Automatic Autonomous Observatory: Application to TAROT
M. Bringer, M. Boër
Centre d'Etudes Spatiale des Rayonnements (CESR/CNRS),
9 av du Colonel Roche,
31028 Toulouse cedex 04, France
F. Morand
Observatoire de la Côte d'Azur, 2130 route de
l'observatoire, Caussols, 06460 Saint Vallier de Thiey, France
Abstract:
We have developed a data processing pipeline for the automatic
TAROT
observatory. It is composed of a scheduler (MAJORDOME), a telescope
control and a data processing software (TAITAR). Details of the
MAJORDOME and classification softwares are given in companion
contributions (Boër et al. 2000; Bringer & Boër 2000). In
this paper, we present the overall architecture of the pipeline, and the
interactions between the various modules.
The Télescope à Action Rapide pour les Objets Transitoires (TAROT,
Boër et al. 1999), in operation at the Calern Observatory (France) has
for primary objective the realtime detection of Cosmic Gamma-Ray Bursts
(hereafter GRBs). However, this goal uses only about 10% of the
available time, hence several routine programs, mainly connected with
the study of celestial variability are currently running.
The users for the routine program submit remotely their observation
requests to the MAJORDOME. If selected by the scheduler, the request
will be processed by the telescope and the corresponding frame will
then go through the data processing software in near real time in order
to generate a catalogue of the different objects including their
properties. If an alert occurs (e.g., to observe a GRB source location
given by the GCN),
the corresponding position is immediately sent to the telescope, and
the images are processed straight afterwards, delaying the processing
of routine observations. With this software suite, TAROT is now fully
autonomous, i.e., there is no human intervention on the telescope, nor
on the processing. Results may then be used to search for new or
variables objects, as it is the case for optical counterparts of GRBs.
The TAROT observatory (Calern, France) running for a year now is
composed of different modules linked to each other according to Figure 1.
Figure 1:
Overview of the TAROT pipeline modules.
 |
- The OBSERVATION REQUEST is the module that enables anyone
to send an observation proposal to TAROT. The request will then
join the request database.
- The GCN-BACODINE module is simply a TCP-IP connection with the
NASA center that is collecting all the satellite GRB event
positions and sends those in real time to automatic observatories
(Barthelmy 1999).
- The MAJORDOME is a software that is able to
optimally schedule observations considering different parameters
such as the Moon position, the transit date and other
constraints eventually given by the OBSERVATION REQUEST. One of the
particularity of TAROT is its reactivity to unprogrammed
events. The MAJORDOME is though able to consider what we call an
alert request (GRB in our case) and to
reschedule in a short time (less than a minute) the timetable for
the rest of the night.
- The CONTROL is the module in charge of the telescope drives and
the housekeeping. It is able to detect rain, dew, and
other conditions that need a special action. It then inform the
MAJORDOME that reacts to these events.
- The DATA PROCESSING software is composed of TAITAR, the image
processing software that detects the objects on the frame, as well
as new or variable stars. We have also developed a cloud algorithm
based on flux measurement of reference sources evenly spaced over the
sky. The last feature of the DATA PROCESSING software is an automatic
satellite detection based on the Hough Transform.
Figure 2:
TAROT receiving an alert from GCN.
 |
The MAJORDOME receives the GRB location through the GCN connection and
instantly stops what it
was doing. He generates another timetable for the rest of the night
and starts sending to the CONTROL the requested positions. The CONTROL
does quite a basic work. Whenever it receives an order, it executes
it, hence the system intelligence lies in the
MAJORDOME
Let's go through our pipeline with the 1st Frame.
As soon as the first frame is written on the hard disk, TAITAR starts
analyzing it. First of all, TAITAR evaluates if there were clouds on
the frame. If it is the case, the frame is not completely analyzed and
the CONTROL and the MAJORDOME are informed of partially sky coverage.
The MAJORDOME can decide to launch a complete cloud detection procedure, in
order to select regions of the request database whose coordinates
correspond to clear sky areas.
If the Frame seems normal, the analysis starts.
The complete analysis of a frame is done in 4 steps (Irwin 1985, Bertin
& Arnouts 1996): estimation of the Sky Background, thresholding,
deblending and finally astrometric and photometric calibration.
At the end of the analysis, TAITAR generates two output
catalogues. One that contains all the sources of the frame, and
another that contains the objects that weren't on the reference
catalogues. Those objects are possible optical counterparts of GRB or
variables objects.
Our system is completely autonomous and doesn't need any human intervention excepted once a week in order to
change the archive DAT.
Our Scheduler is able to deal with a request database of 1500 requests in less than a minute. This is done every
day at noon for the next night, or after each interruption due to rain or cloud detection. Table 1 gives an
example of the various possible requests, as well as the results of the MAJORDOME scheduling.
The efficiency 
of the system is defined as the ratio of the
effective observing time by the total night time (Here 6h 19 06).
For our example here, we have obtained 
.
TAITAR fully process a 1280x1024 frame in 90 seconds. The extraction of objects lasts 15 to 20 seconds. The rest of
the time is used to calibrate the frame with the USNO catalogue. We are looking forward to
generating our own catalogue in order to optimize the astrometric and
photometric calibration.
Astrometric Accuracy: Our objects location average error is 1 arcsec.
Photometric Accuracy: Our differential magnitude accuracy is
0.03 mag.
TAROT produces about 300 frames per night, which represents
. Our system analyzes these
frames in 7.5 hours.
We have developed a fully autonomous pipeline in order to run the TAROT
observatory. Our system is composed of different modules that interact with
each other. We are now looking forward to developing a second version of
our scheduler that will be able to schedule observations even in bad
conditions and on a period of several months. Shortly, we will use on
line our classification algorithm based on Kohonen networks and our
satellite detection algorithms which detect in less than 10 seconds all
the satellite or plane tracks on the frame.
Acknowledgments
The TAROT experiment has been built with the support of the Centre National de la Recherche Scientifique,
Institut National des Sciences de l'Univers (CNRS / INSU).
References
Barthelmy, S. 1999, information available at http://gcn.gsfc.nasa.gov/gcn/
Bertin, E. & Arnouts, S. 1996, ApJS, 117, 393
Boër et al. 1999, A&AS, 138, 579
Boër, M. et al. 2000, this volume, 115
Bringer, M. & Boër, M. 2000, this volume, 640
Irwin, J. 1985, MNRAS, 214, 575
© Copyright 2000 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: Observation Interval Determination for the Chandra X-ray Observatory
Up: Data Pipelines and Quality Control
Previous: Generating Calibration Reference Files with an OPUS Pipeline
Table of Contents -
Subject Index -
Author Index -
PS reprint -
adass@cfht.hawaii.edu