The Fairborn/TSU Robotic Telescope Operations Model

Robotic Telescopes
ASP Conference Series, Vol. 79, 1995
Gregory W. Henry and Joel A. Eaton (eds.)

Gregory W. Henry

Center of Excellence in Information Systems, Tennessee State University,
330 10th Avenue North, Nashville, TN 37203

Abstract. The operation of multiple robotic telescopes at Tennessee State University is described. These telescopes are all located at a single site, and each telescope is dedicated to a single, long-term observing program. These features result in an automated data acquisition, reduction, and archiving system that is extremely productive, yet inexpensive to operate and easy to manage.

1. Introduction

For the past several years, Tennessee State University has operated three telescopes for automated photoelectric photometry at Fairborn Observatory's APT site in southern Arizona (Genet et al. 1987; Genet, Boyd, & Baliunas 1986). Fairborn Observatory is a non-profit organization headed by Lou Boyd for the development and operation of robotic telescopes. Its observing site is located at the Smithsonian Institution's Fred L. Whipple Observatory on Mt. Hopkins in the roll-off-roof enclosure of the Smithsonian's old satellite tracking installation at 7800 feet above sea level (Burke & Kirchhoff 1968). The first of the three telescopes to go into operation (in early 1986) was the Fairborn 10-inch. By the end of 1994, this telescope had collected over 49,000 group observations of variable stars, the majority of which are semi-regular variables in a collaborative program with the Harvard-Smithsonian Center for Astrophysics. These group observations include all comparison star, check star, sky, and dark integrations in all desired filters needed to determine differential magnitudes for a given program and check star. The second telescope, a 16-inch, began operating in November 1987 and has collected 72,000 group observations of chromospherically active single and binary stars in a collaboration with Vanderbilt University. The third is a 30-inch telescope that has made 11,000 high-precision observations of solar-type stars since April 1993, also in collaboration with the Center for Astrophysics. These three telescopes and the precision of the data acquired with them are described in my paper in Part I of this volume. A fourth telescope, a 32-inch dedicated to high-precision photometry of solar duplicate stars, will go into operation in early 1995.

2. Description of the Operations Model

Two primary features that define the Fairborn/TSU robotic telescope operations model are (1) all telescopes are located at a single site and (2) each telescope is dedicated to a single long-term observing program. Our operation uses only one person (Boyd) at the remote site to develop, maintain, and oversee the operation of the telescopes and one person (Henry) at Tennessee State University who is responsible for scheduling the observations on the telescopes, performing daily quality control checks, reducing and archiving the data, and troubleshooting problems with telescope performance. We have designated these two positions the Principle Engineer (PE) and Principle Astronomer (PA), respectively. Scientific collaborators are also essential to help digest the tremendous flow of data that results from this operation, which we feel is the simplest and most effective way to operate small robotic telescopes.

There are several advantages to having all telescopes located at the same remote site and within a single enclosure. Physical plant requirements such as road access, power (and power backup systems), phones, computer network connections, site maintenance and security are all greatly simplified when only one site needs to be supported. A single set of weather sensors and a single site control computer for monitoring weather conditions and for opening and closing the roof can service all of the telescopes and protect them when the weather deteriorates. A single communications computer can forward observing requests from the PA to the appropriate telescopes, gather the nightly data from each telescope, and forward them to the PA each morning over the Internet. Workshops, tools, and electronic test equipment can be can be easily available on site. Instrumental support systems such power supplies, air filtration and drying equipment, and temperature control systems can be centralized and support all telescopes. Perhaps the most significant advantage is the ability of a single PE to manage the site in support of all telescopes. In fact, this has been done by Boyd for a decade in his spare time!

Additional advantages arise when each telescope managed by the PA is dedicated to a specific, long-term observing program. Each program can benefit from large quantities of high-quality data taken year-round from an excellent observing site. Because these data are taken with dedicated instruments, they benefit from an internal consistency not possible with traditional manual observing runs on multi-user telescopes. Standardized observing sequences on each telescope that are optimized for each project make it simple to program the telescopes to make the observations but, more importantly, make data reductions straight-forward andefficient since automated reduction and archiving programs can be written that require very little effort by the PA to oversee. Also, standard star and other quality control observations tailored to the specific observing program can be developed to ensure that the data are of the highest possible quality, and the reduction of these observations can also be automated. With only one scientific program running on each telescope, complete sets of standard star and quality control observations can be made each night on each telescope while sacrificing only a few percent of the observing time. With the automated reduction routines, a PA can process the nightly output of several telescopes in only a few minutes each day. Additional advantages in terms of cost and quantity of data produced are reviewed below.

There are admittedly a few disadvantages of our robotic telescope operations model, and there are several places where the operation is subject to single-point failure. Obviously, if it is cloudy over Mt. Hopkins, we will not get data from any of our telescopes. We do not benefit from a distribution of our telescopes in longitude around the earth or in complementary weather patterns at multiple sites. Power failures, ice storms, failure of the site control computer to open the roof, failure of centralized cooling systems, etc. all can similarly result in the total shutdown of our system. A couple other possible sources of single-point failure are simply too disquieting to mention. In addition, telescopes dedicated to long-term programs cannot be very responsive to short-term or target-of opportunity projects without some sacrifice of the long-term projects. On the positive side, however, our telescopes have been up and operating for the past several years on well over 95% of the available clear nights on Mt. Hopkins, and our observational programs have benefited tremendously as a result. Tennessee State University recently made an attempt to overcome the difficulties inherent in a multi-site operation and eliminate the single-point failure modes by placing a 32-inch APT on Mt. Wilson, but the failure of AutoScope Corporation to produce a working instrument forced us to retreat to our single-site operation. For now, this seems to be the only practical option available to us.

It can also be mentioned that other robotic telescopes at the Fairborn site are operated in modes other than our own. One 30-inch APT is owned and operated by a consortium of four universities, and the observing time is shared among several astronomers, one of whom serves as PA for the telescope (Dukes et al., this volume). Two additional 30-inch and another 10-inch APT are owned by Fairborn and operated in a "rent-a-star" mode whereby observing time is purchased by a large number ofobservers on a per-observation basis, again with a single PA for each telescope to oversee the scheduling of observations and distribution of data (Seeds, this volume). Boyd also serves as PE for these telescopes.

3. The ATIS Standard Programming Language

An implicit part of the operations model described here is the language in which the individual robotic telescopes are programmed. In 1989, Fairborn Observatory developed the Automatic Telescope Instruction Set (ATIS) to define the observational requests for the Fairborn APT's (Genet & Hayes 1989). For the first time, this new language allowed APT users to tailor the exact specifics of their observing sequences, to communicate them to the telescopes via ASCII file transfers over telephone lines or the Internet, and to retrieve the resulting observations the next day. In ATIS, a group observation is the primitive unit to be scheduled and executed. The ATIS groups consist of sequences of telescope and instrument commands composed by an astronomer to accomplish a given observation and contain commands to move the telescope, to acquire and center stars, to control the filters, and to make integrations in a specified sequence.

In addition to specifying the syntax and semantics for observation requests and results, the ATIS language provides a set of group selection rules that are used to determine the execution order of groups during the night. The group selection rules provided by ATIS essentially implement a first-to-set-in-the-west policy: at any given point in time the telescope observes the star that will next move out of the hour angle limits in which it can be observed. Through the use of user-input parameters to define group types, group priorities, probabilities of execution, date and local sidereal time limits, number of observations requested, etc., ATIS allows APT users to define any desired observation sequence as well as to intersperse standard star observations and other quality control checks throughout the night and have them all scheduled automatically. These group selection rules allow our robotic telescopes to run for long periods of time, automatically selecting targets to observe each night in the face of changing seasonal availability of stars and interruptions due to bad weather.

However, in spite of the extraordinary improvements to the long-term scientific programs being conducted at Tennessee State University and elsewhere that have resulted from the use ofautomatic telescopes running the ATIS control language, the 1989 version of ATIS is not without its limitations. For instance, there is no look-back or look-ahead capability that might help the telescope decide when to make a given observation, perhaps to fill in a variable star's light curve automatically or make an observation that got missed on a previous night. Many situations can arise, particularly on multi-user telescopes, when more intelligent scheduling than the simple ATIS group selection rules can provide is needed so that the highest priority observations are made at scientifically appropriate times and all users are treated fairly in the allocation of telescope time. Also, since ATIS only allows access to the telescope before the beginning of the night and after the end of the night, no information is available during the night about the current status of the observing program, the quality of the data being obtained, or about the performance of the telescope and detector that might allow appropriate modifications of the observing program to be made in real time.

To address these limitations of the original ATIS control language, Fairborn Observatory collaborated with a committee of APT users and with the artificial intelligence (AI) group at the NASA Ames Research Center to develop a new version of ATIS with the capability to override default ATIS group selection rules. A mechanism was developed for communication with a telescope during the night using incremental ATIS partial input and partial output files. This new feature makes it possible to implement an external AI scheduler (running on a remote machine) that can effectively drive the telescope controller in near real time to improve the scheduling of the observations. The result, including additional enhancements to ATIS, was published as ATIS93 by Boyd et al. (1993) in hopes of its becoming an international standard language for automatic telescope control. The article by Drummond et al. in this volume describes how ATIS93 will be incorporated into a new set of software tools called the Associate Principle Astronomer (APA) being developed at NASA Ames that will include an AI scheduler and greatly facilitate the management of robotic telescopes.

4. Comparison with the Old Manual Operations Model

It is interesting to compare our robotic telescope operations model with previous manual observations in terms of the level-of-effort and costs per observation. For the manual observing discussion, I examined old observing logs from 1980, a year when I was fully employed by Vanderbilt University and my first prioritywas to make manual photometric observations of chromospherically active stars. During that year I had essentially unrestricted use of the Dyer Observatory 24-inch telescope to use whenever it was clear. Nashville is not the most promising site from which to do photometry; my observing log shows that I observed at Dyer on 40 different nights during 1980 and made a total of 396 complete differential group observations. All 40 nights were not clear all night; the number of observations per night ranged from 1 (when it became non-photometric shortly after I began) to 28 (when it was clear all night long). The average number of observations per night was 10.

Because of the relatively poor observing prospects in Nashville, I applied for two observing runs on Kitt Peak with the No. 4 16-inch telescope. Of the 42 nights I received in 1980, I was able to observe on 34 and made a total of 601 group observations. The number of observations per night ranged from 1 to 32 with an average of 18. The higher average compared to Dyer is expected since the chances for a full night of observing at Kitt Peak were higher than at Dyer. The actual efficiency of manual observing with the two telescopes was comparable; full nights of observing resulted in approximately the same number of observations from each telescope. However, it was apparent that a few weeks of observing time at a remote but good site resulted in more observations than could be acquired during the rest of the year from the home site located in less favorable observing conditions (601 and 396, respectively).

In sum, I observed on 74 nights in 1980 and obtained 997 group observations. I estimate that I spent approximately 50% of my time obtaining these observations. This included preparation of proposals to Kitt Peak, travel, observing, preparation of finding charts, wasted effort on nights that turned non-photometric, data reduction (which in those days was nearly as much drudgery as the actual observing), and recuperation. Therefore, roughly 1000 differential group observations per year were made and reduced by an observer working approximately half-time from a typical eastern site with occasional trips to a remote site while using typical manual telescopes and data logging systems common 15 or 20 years ago.

For comparison with automated observing, I examined a recent 12-month interval when the 10-inch, 16-inch, and 30-inch APT's were fully operational. During this time, the Fairborn 10-inch secured 7427 group observations, the Vanderbilt/Tennessee State 16-inch made 13,952 group observations, and theSmithsonian/Tennessee State 30-inch obtained 7246 observations. The total number of group observations from all three telescopes during this 12-month interval was 28,625.

However, the typical APT group observation was more complex than the typical manual group observation from 1980. In particular, more check and comparison stars were used, observations were made in more filters, and many more standard star and quality control observations were made with the APT's. Indeed, the average APT group observation represents at least twice the data as the average manual group observation. So a very conservative comparison reveals that our robotic telescopes produce fifty times more data per year than one of my best years of manual observing. This will increase even further when our 32-inch APT goes into operation in 1995.

Once the automated telescopes are built, installed, debugged, and working smoothly, and the long-term observing programs are set up to run on them, and the automated data reduction, quality contol checks, and archiving software are written, very little time is required on a daily basis to manage the telescopes. I certainly spend less than 10% of my time on this and probably considerably less than that during normal weeks. Boyd estimates that he spends at most a few percent of his time on maintenance of these three telescopes. So, again being very conservative, we can say that in one-fifth of the time spent, we make 50 times more observations with our robotic telescopes than was possible with previous manual techniques. This corresponds to a 250-fold increase in data acquired per unit time spent by the astronomer.

One caveat should be expressed here. The kind differential photometric observing I have been discussing involves hundreds of moves between stars each night with relatively short (10 - 20 sec) integration times on each star. Moving the telescope, identifying the proper star, centering the star, moving filters, starting integrations, and logging data are all operations an automated system can do much more efficiently than a manual observer. However, for observing programs involving much longer exposure times (such as high-resolution spectroscopy), a manual observer could work nearly as efficiently as an automated system, at least until he fell asleep. So not all kinds of observing will show the same gains in observing efficiency as have been achieved with automated differential photometry. Even so, there are still many good reasons for automating other kinds of observing programs (Eaton, this volume).

Finally, we can now compute a cost comparison between our automatic observing system and earlier manual observing. For the 12-month period cited here, the 16-inch APT, for instance, made 13,952 group observations. Fairborn operates and maintains this telescope for $15,000 per year. Therefore, the cost per observation was $1.07 per group. To operate a 16-inch telescope manually on every clear night of the year, (at least) two full-time observers would be needed. At $30,000 per year each plus benefits and overhead, this would cost $90,000 per year. On a typical good night, the 16-inch APT makes approximately 100 group observations. On my very best nights of manual observing, I could make only one-third that number, and each manual group observation contained only one-half the data. Therefore, the manual observers will get one-sixth the number of observations per year, compared to the APT, and at six times the cost. Therefore, the cost per observation obtained manually would be 36 times the cost of an APT observation.

All of these considerations, combined with the order of magnitude improvement in the precision of APT observations (Henry, this volume, Part I) dramatically illustrate that the operation of automated telescopes for dedicated, long-term differential photometric observing programs (i.e. the Fairborn/TSU model) has brought revolutionary advances to this field.

Acknowledgments. The development and operation of robotic telescopes and the analysis of data from them has been supported for several years at Tennessee State University by the National Aeronautics and Space Administration and by the National Science Foundation, most recently through NASA grant NAG8-1014 and NSF grant HRD-9104484. None of this work would have been possible without the dedicated efforts of Lou Boyd at Fairborn Observatory.


Boyd, L., Epand, D., Bresina, J., Drummond, M., Swanson, K., Crawford, D., Genet, D., Genet, R., Henry, G., McCook, G., Neely, W., Schmidtke, P., Smith, D., & Trueblood, M. 1993, IAPPP Comm. No. 52, p. 23

Burke, J. J. & Kirchhoff, W. 1968, Sky & Tel., 36, 284

Genet, R. M., Boyd, L. J., & Baliunas, S. L. 1986, IAPPP Comm. No. 25, 15

Genet, R. M., Boyd, L. J., Kissell, K. E., Crawford, D. L., Hall, D. S., Hayes, D. S., & Baliunas, S. L. 1987, PASP, 99, 660

Genet, R. M. & Hayes, D. S. 1989, Robotic Observatories (Mesa: AutoScope)

Henry, G. W. 1995, IAPPP Comm. No. 57, p. 74

Seeds, M. A. 1994, IAPPP Comm. No. 56, p. 23