 |
With the advent of automatic telescopes and the use in service mode of large facilities, the problem of efficient night scheduling is becoming more important. In the usual configuration, the observer is allocated a given amount of nights. A mixed approach has been introduced for the ESO NTT and VLT telescopes, with the use of the concept of observation blocks (Chavan et al. 1997). These blocks are handled by three schedulers, in charge respectively of the long, mid- and short terms. Bresina (1998) defines individual groups of observations sent to the telescope via a centralized system, the Principal Astronomer, with pre-affected priorities and time constrains. The system chooses the best sequence which optimizes an objective function.
The first version of our MAJORDOME software (Bringer et al. 2000a) was written for the use of the Télescope à Action Rapide pour les Objets Transitoires (TAROT - Rapid Action Telescope for Transient Objects; Boër et al. 1999), a fully automatic instrument in operation at the Calern Observatory (France). The primary goal of this instrument is the rapid detection and follow-up of Cosmic Gamma-Ray burst sources (hereafter GRBs) at visible wavelengths. For that purpose TAROT is able to slew to any location over the sky in less than 3 seconds (usually 1-2s), upon reception of a position sent by the GRB Coordinate Network (GCN) ), few seconds after the GRB event has been detected. Since GRB observations uses only 10% of the telescope time, we use TAROT for various scientific objectives, usually connected with celestial variability. In order to accommodate all the above mentioned objectives, and to observe as efficiently as possible, we developed a suite of original algorithms and softwares, the MAJORDOME, in charge of the telescope schedule, which takes over the CONTROL software in the absence of any safety alarms (for a general view of the TAROT pipeline, see Bringer et al. 2000b).
Below we list some terms and concepts used for the scheduling by the MAJORDOME:
Following these principles we designed the MAJORDOME, the software which is already in use by TAROT. The time line is computed every day for the next night, using all the visible requests from the database. As soon as a group has been validated, the corresponding request is removed from the database, if not periodic. This software schedules observations with an efficiency above 90%. However, we encountered some difficulties handling periodic or constrained requests with a large time interval, and it is unable to give to the telescope users and operators the visibility over the schedule. We though also that a larger programming horizon would result in more efficient optimization. The following guidelines are used:
Figure 1 displays how the time line is built as a sliding, daily computed, series of mid-term time lines. The requests may arrive at any time. At a given time, the new nightly schedules are computed for a week. If an unexpected event happens (TOO, interruption...), a new schedule is computed for the remaining of the night. The normal process resumes the day after, including the un-observed scheduled requests. Figure 2 displays a functional diagram showing the reactivity of the system to unexpected events..
We added also a ``filter'', able to decide in which class will be an observation, and to reject eventually incorrect requests, or to react in case a given class (e.g. periodic or constraint) is overloaded. The MAJORDOME will also allow certain requests to be processed before the astronomical twilight or dawn (e.g. to search for comets or space debris), or in bad conditions (like partial cloud coverage, non-photometric night...). We are currently implementing the algorithms needed for this more complicated approach.
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).
Boër M., et al. 1999, A&AS, 138, 579
Bresina J. 1998, Ph.D. thesis, Rutgers
University, available at
http://ic-www.arc.nasa.gov/ic/projects/xfr/sampling/thesis/
Bringer, M., Boër, M., Peignot, C., Fontan, G., & Merce, C. 2000a, in preparation
Bringer, M., Boër, M., & Morand, F. 2000b, this volume, 445
Chavan et al. 1997, SPIE proceedings, 3349, 97
