All the major components of the United Kingdom Infra-Red Telescope (UKIRT) Upgrades program are now in place. The thrust of the program has shifted to developing the new telescope capabilities so that performance is maintained under real observing conditions. This paper presents an overview of the current state of affairs and focuses on how we have implemented the secondary mirror and fast guide systems and how we control the active primary mirror system to minimize the telescope aberrations.

The NTT project had the principal aim of field testing the VLT control system prior to installation on UT1 on Paranal. In July 1996 we began installing the control electronics and software. The telescope was stripped down to the field electronics and completely rewired. First light was achieved 2 months later and the integration of the system was completed 4 months after that. The VLT control system has been proven to be functional and fundamentally sound. Over 80% of the VLT control system is mirrored on the NTT with deviations only allowed where the hardware made it impossible to reuse the code prepared for Paranal. The NTT will operate with the complete control system executing service observations following the shake down period in early 1997.

The very large telescope (VLT) software commissioning has started since a while, earlier than any VLT subsystems were ready for integration. This was possible thanks to the new technology telescope (NTT) upgrade (reported in a separate paper, Ref. 3), which shares most of the software with the VLT. The integration tests with the main VLT structure do also represent another fundamental milestone in the software commissioning process (see also separate paper, Ref. 4). The whole control software is based on a very distributed computer architecture. The final layout of the computers (work-stations and microprocessors), networking devices and underlying concepts have been tested both at the NTT and on the so called VLT computer control model, a relevant off-line subset of the computer equipment to be used in the VLT control room and telescope area for one unit telescope. The VLT common software including a real-time database, is the stable core of the whole VLT control software. This comprises also high level applications, like the real-time display (RTD), the panel editor and the CCD software to be used for technical CCDs. It is distributed with a policy of regular releases, subject to automatic regression tests and is used by VLT Consortia and contractors too. New modules have been added to insert the VLT control software in the data flow context, interfacing it in particular to schedular and archive. The VLT software support team will soon start regular operation at the VLT Paranal site, providing continuity between the integration activities of the various subsystems. They will be the front-end of the commissioning effort, based also on background support provided over links from the European headquarters.

This paper describes the selection process for the encoding system of the main (azimuth and elevation) axes of the Gemini 8-m telescopes. The main part of this is the description and results of lab tests carried out on competing systems. Two tape encoder solutions existed, one based on an inductive tape and one based on an optical tape, but neither met all the requirements. The optical system provided the better error performance, but its robustness was questionable and it was difficult to interface to the rest of the Gemini servo hardware. The inductive system was already proven to be robust and provided a standard quadrature encoder output. However, the system didn't meet some of the performance requirements and was not supplied as a complete system. The lab tests were carried out to try and resolve some of these problems and to help arrive at a decision between the two systems. The results showed that the error performance of both systems, in the presence of compensation, was good enough for the telescope application. The final decision was based upon a formal tender exercise that compared many properties of each system.

Inexpensive optical rotary incremental encoders now available can provide resolution approaching one arcsecond. However, several factors limit the accuracy of measurement that can be obtained. We report on test results of rotary incremental encoders obtained with a test setup that compared the output of two such encoders driven by the same shaft. Although intrinsic non-linear response of the encoders tested is specified to be less than plus or minus 15 arcseconds, additional errors are often caused by the coupling of the encoder to a rotating device. Bearing runout and shaft misalignment typically require use of a flexible coupler, but tests with several types of small inexpensive flexible couplers have shown that these can contribute additional errors including windup and non-uniform rotation that is affected by small changes in alignment. An additional minor source of error is due to a reproducible periodic error of several arcseconds generated in the interpolation electronics used to provide high resolution by subdividing the analog signal from the encoder. The driving torque required by a typical Gurley encoder is larger than might be expected, and has been measured at various speeds by determining the amount of windup with a solid aluminum coupling shaft.

Subaru observation software system (SOSS) provides observers with so called high-level user interface, scheduler and data archival system. This paper presents both software and hardware environment for test, debug and development of instrument controller (OBCP) and SOSS. The environment is composed of instrumentation software toolkit, instrumentation software simulator, telescope simulator and summit simulation computer system.

The data flow system (DFS) for the ESO VLT provides a global system approach to the flow of science related data in the VLT environment. It includes components for preparation and scheduling of observations, archiving of data, pipeline data reduction and quality control. Standardized data structures serve as carriers for the exchange of information units between the DFS subsystems and VLT users and operators. Prototypes of the system were installed and tested at the New Technology Telescope. They helped us to clarify the astronomical requirements and check the new concepts introduced to meet the ambitious goals of the VLT. The experience gained from these tests is discussed.

Many telescope control systems now make use of the so-called 'virtual telescope' concept -- a software abstraction of the real telescope which masks imperfections in the hardware from higher levels of the software. In general, this approach allows for elegant and rigorous control of telescope pointing and tracking. When slewing, however, while the virtual telescope arrives on source immediately, the real telescope only catches up after some time. This is especially a problem when performing raster-scanned observations: since the demand position and velocity have discontinuities at the end of each row, a naive implementation of the standard virtual telescope system results in missing the demand positions at the start of each row. In the existing JCMT telescope control system (TCS), this problem is solved by having the TCS calculate a route in (az,el) space for the real telescope to follow which results in it arriving at the correct position, moving with the correct velocity, at a predictable time in the future. In this paper we describe a generalized implementation of this technique, which has the added advantage that the 'astrometric kernel' and 'telescope servo' layers are cleanly separated, allowing telescope-specific hardware to be combined with a generic astrometric kernel. Since the solution requires only minimal changes to the standard virtual telescope design, this approach may be of interest to other telescopes which are currently using, or are planning to use that design.

We describe the design model and implementation of the data archiving system used by the Berkeley-Illinois-Maryland Association (BIMA) millimeter array telescope. This system transmits data in real-time from the BIMA telescope at the Hat Creek Observatory in northern California via the Internet to an archive server at the National Center for Supercomputing Applications (NCSA) in Urbana, Illinois. Once at NCSA, the data undergo minor processing to make it available to astronomers via an HTML interface: (1) the data are checked for successful transmission, (2) metadata are extracted from the data and inserted into a searchable database, and (3) the data are sent to the NCSA mass storage system. When necessary, the system can carry out rollback operations which allow it to easily recover from errors, particularly those associated with the often unstable Internet. We also comment on some ways in which the system can be improved and expanded to adapt to changing observing strategies.

We describe the remote observing capabilities currently provided at the Keck Headquarters in Waimea (located approximately 32 km from the Keck summit and at an elevation of 850 m) as well as the subset of capabilities now available from the mainland via the Internet. The bandwidth available between the telescope and the remote observing site determines which of several remote observing software and networking architectures is most cost-effective. We describe our operational experience with several different architectures, differentiating between those used at Keck headquarters and those used for remote observing from the mainland. Methods for optimizing bandwidth are explored, including the pipelining of image readout with data compression and transmission to the remote site. Tradeoffs between network bandwidth, security, and portability of software to remote observing sites are also explored.

In order to achieve diffraction-limited performance in a distributed aperture optical system such as the multiple mirror telescope, the elements must be optically phased to form a single aperture which can be a difficult task. In this paper, we report a novel, iterative adaptive control method, called far-field optimization, which employs the simplex algorithm to configure the elements of a distributed aperture telescope to optical alignment. Far-field optimization does not require the knowledge of the adaptive mirror surfaces, thus eliminating the need for a wavefront sensor, but uses a simple measure of the point-spread function (image-plane intensity). We present results of computer simulations to demonstrate the utility of far-field optimization to remotely align the six-element multiple mirror telescope in a few seconds, even in the presence of drifting atmospheric turbulence.

The portable telescope control system (PTCS) project is a collaboration between the Anglo-Australian Observatory and the Joint Astronomy Centre, Hawaii. The project aims to develop telescope control software in a form which is portable between a wide range of computer systems, and which can easily be adapted to different telescopes. PTCS uses the DRAMA software environment which provides a high degree of operating system independence. The PTCS design is based on tested algorithms used in existing successful telescope control systems. The initial version of PTCS is now being interfaced to an EPICS based drive servo system for use with the James Clerk Maxwell sub-millimeter telescope on Mauna Kea.

In the latest generation of astronomical telescopes the increase of the primary mirror diameter has placed ever increasing demands on the technical performance of the mirror support systems. In this paper the authors discuss the mechanical and the electronic active control system design and subsequent tests of the position actuator prototype that mechanically link the 8.4 m honeycomb mirrors of the large binocular telescope to six rigidly reinforced locations in each primary mirror cell structure. During telescope operation, these adjustable length actuators precisely control the six degrees of freedom of motion of the mirror. Each actuator has a high mechanical axial stiffness and, as new feature, an active control system, based on piezoelectric elements, in order to increase its axial stiffness, with a bandwidth from dc up to 30 Hz, assuring that the natural frequencies of the mirror do not degrade the optical performance of the telescope. Moreover, other requirements have been satisfied in the mechanic of the actuators: flexures are provided on each end to minimize any moments applied to the attachment of the actuator to the mirror; one axial load cell for each actuator provides a precise measurement of the external forces applied to the mirror, such as wind loads, for the control of the pneumatic force system that supports the weight of the mirror; a very sensitive and precise capacitive sensor measures the total length of the actuator to sub-micron resolution upon request. Each actuator has a reliable fail- safe system that limits the compressive and tensile forces that can be applied to the mirror. The mechanical and the electronic design, and the later experimental tests, of this actuator prototype have been performed in the Arcetri Laboratories under the supervision of the authors of this paper.

The installation of the complete VLT telescope control system on the observatory is a complex task. It is important that the various components of the system have been carefully tested and integrated before. This paper presents the ESO strategy to pre-installation testing. In particular, results and experiences from pre-erection tests of the telescope structure are presented. In these tests, the complete telescope structure, including both axes with encoders and drives, has been built up at the premises of the European manufacturer (in Milan, Italy). These tests provide valuable input for the erection on Paranal. To this system, ESO added control electronics and software, which was tested with the telescope. The complete positioning of both main axes is under test, including slewing and tracking performance tests, as far as this is possible without using the sky. The VLT control software and most parts of the VLT control electronics have also been tested on the NTT on La Silla. Since the NTT upgrade software is practically a subset of the VLT software, the NTT tests have provided invaluable feedback for the VLT. The NTT tests are described in a separate paper presented at this conference. The paper also briefly discusses subsystem tests, and presents results from some of the subsystem tests performed in Europe.

A discussion of the interactive operator user interface developed for the Gemini 8-m Telescopes is presented. Topics include the use of a layered synthesized view of the area of interest on the sky, a data driven approach to the control of the subsystems, and an adaptive view on the health of those subsystems. The synthesized view utilizes information from pre-existing databases; guide, wavefront and scientific detector image data; as well as operating and performance limits. This information is presented to the user in layers, with each layer containing an observer, subsystem or other logically oriented view. The ability to control which layers are presented, as well as which parameters are directly modifiable is vested in both the user and the configuration software. Implementing this above a data driven control interface encourages the use of observing templates. Exhaustive parameter control with parallel realization in the lower level mechanisms results in fast, fine grained and repeatable control. Combining the major control interfaces, each with a different view of the desired behavior, error in behavior, and possible corrections allows the operator to spend more time optimizing observations, rather than setting up equipment. Maximizing time with quality light falling on the science detectors is a primary goal.

The Keck 1 chopping secondary was built by the Palo Alto Research Laboratories of the Lockheed (now Lockheed Martin) Missiles and Space Company. The only software component of the delivered system is a proprietary error correction algorithm; Keck wrote software to generate acceleration-limited azimuth and elevation demands, to rotate these demands as a function of telescope position, to interact with the error correction system, and to mange hardware start-up and shutdown. The Keck 2 chopping secondary, also built by Lockheed, was originally conceived as an infrared fast steering mechanism (IFSM) and is simpler than the Keck 1 system, with lower power and acceleration limits and, therefore, lower chop amplitude and frequency specifications. As far as possible, it provides the same external interfaces as the Keck 1 system. A new EPICS- based telescope control system has been written for Keck 2 and was retrofitted on Keck 1 in March 1997. The Keck 1 chopper control software has been converted to the EPICS environment and, at the same time, altered so that the same software supports both choppers. This conversion has retained as much as possible of the complex real-time code of the old system while at the same time fully utilizing EPICS facilities. The paper presents more details of both the old and the new systems and illustrates how the new system is simpler than the old as well as being much better integrated into the overall telescope control system. Operational experience is presented.

In an aplanatic two mirror telescope, spherical and coma aberrations may be introduced if a misalignment of the secondary mirror is present. That misalignment may be intrinsic to telescope, due to small errors in optical design or manufacturing; however it will always be present during the life of the telescope, due to maintenance errors, thermal distortions of the mirror and the structure, mechanical distortions of the mirror mount, mechanical arrangement of the structure and other. An active control of the secondary mirror, allowing free positioning of the secondary, is important to correct such unwanted effects. The Italian Galileo National Telescope is equipped with a secondary mirror supported by an 'hexapod' structure, allowing a complete positioning control. In this paper a strategy for handling the positioning and movement of the exapod support of a secondary mirror will be investigated from two point of view: an analytical and a neural network approach.

The Gemini telescope control system calculates the orientation of the Cassegrain rotator in a way that allows the desired angle to be specified relative to any of the four supported tracking coordinate systems (FK4, FK5, geocentric apparent place and topocentric az/el) independently of the tracking frame of the mount. In addition, the actual, rather than demanded, position of the rotator mechanism is used in the calculation of the mount position so that the telescope is able to track a target with an off-axis pointing origin (By pointing origin we mean the nominated point in the focal plane to which the pointing refers; i.e. where the image of the celestial object being tracked will appear.) even when the rotator is not in position. The actual rotator position is also used in the calculation of guide star coordinates so that guide stars can be acquired before the rotator is in position. This feedback of the actual rotator position to the mount calculation results in some unexpected behavior, particularly near the zenith blind-spot where the mount may track either north or south of the zenith depending on the positions of the rotator and the nominated pointing origin. The algorithm used to calculate the desired rotator position and the measures taken to avoid the unpredictable behavior near the zenith are described.

We describe the control and user interface software for the image rotator on the high resolution echelle spectrometer (HIRES), located at the Nasmyth focus of the Keck-1 10-meter Telescope. This image rotator counteracts the field rotation induced by the alt-az telescope mount and aligns the image delivered to the spectrometer slit and the slit-viewing TV camera. The rotator can align the image so that either a specified position angle on the sky or the atmospheric dispersion vector is held fixed relative to the slit. The rotator is physically an integral part of the spectrometer; but, since it affects telescope pointing and auto-guiding, it can also be considered part of the telescope. Accordingly, it interacts with the control systems for both the telescope and spectrometer. The apportionment of rotator control functions between these two systems poses interesting operational challenges. Since the telescope operator and the observer each need to operate the rotator at various times, their respective user interfaces must provide consistent rotator control and status functions.

The Michelle Edict array control system is being built for use with the Michelle mid-infrared spectrometer/imager on UKIRT and Gemini. It will drive large format arrays, such as the Boeing/Rockwell 256 by 256 BIB and Raytheon/SBRC 320 by 240 IBC hybrid SiAs devices. It provides for rapid real-time processing and export to a host computer, storage and quick- look display of the data. To limit critical heat dissipation and mass, the system uses a minimum of front-end electronics at the telescope linked via digital fiber optics to custom- built PCI mezzanine cards. These are installed on several Heurikon Baja 4700 VME cards in an off telescope enclosure. This distributed architecture has small electrical infrastructure requirements and allows Michelle to be moved quickly between operation on UKIRT and Gemini with little impact on other instruments. The use of VxWorks on the Baja processors and the PCI standard allows the system to be easily ported to other VME processor boards supporting the PCI interface. Alongside a cryostat control system, edict interfaces to the data handling systems and the EPICS-based Gemini telescope control. On UKIRT, it will function under a UNIX-based observation control system that is being built to replace the existing VAX/VMS-based ADAM system.

The Liverpool Telescope will be a fully robotic 2 m telescope situated in La Palma and managed from Liverpool John Moores University. The data flows between the robotic components of the telescope are discussed and the functionality of those components described. We pay particular attention to the design of the telescope scheduler. Using simulations we show that a relatively simple dispatch scheduler may be tuned to produce a schedule that has acceptable trade-offs between conflicting scheduling criteria.

Queue scheduling will be one of the major modes of operation of the Gemini 8 m telescopes in which scientific programs will be carried out on behalf of applicants by Gemini staff. Use of a substantial fraction of the available telescope time in this manner will permit access to the exquisite conditions of image quality and background which the telescopes are designed to exploit as well as matching the demands of individual observations to the current conditions. In a previous presentation (SPIE 2871) the classical scheduling and loading of the queue were described. In this paper we discuss the philosophy and parameters which define its execution. Results from detailed simulations of the queue execution process are presented.

The new 8-meter class telescopes represent large investments by the development communities. This means that these telescopes must be operated efficiently to provide the best possible return on these investments and a great deal of effort has been made to provide control software that supports effective use of the telescopes. However, efficient use must be more than just keeping the telescopes operating; it is important that observers be provided tools that enable them work effectively. The Gemini 8 m Telescopes have developed a strategy for helping astronomers plan observations through the design of science programs. While there are a number of unique aspects to this strategy, this paper focuses on the methods used as the foundation for connecting astronomers to the facilities of the observatories during the design of science programs. The methods under development take advantage of emerging Internet technologies to help reduce the maintenance issues normally associated with supporting remote sites, while freeing users from many of the performance problems associated with web-based solutions.

We discuss the requirements and design of the communications system for the Gemini 8-meter Telescopes Project. This system is unique not only in our integrated approach to data, voice and video, but it also must span the globe to reach the two telescope sites in Chile and Hawaii, and provide access to astronomers around the world. We discuss the various services planned for the communications system, the many locations which must be served, and the anticipated quality-of-service demands. The constraints which limit our options are also discussed. Finally we present our plans for meeting these challenges.

Control software of FOCAS (faint object camera and spectrograph) is being developed as a prototype of control software for SUBARU observing instruments. The software system consists of several processes; a network interface process, a user interface process, a central control engine process, a command dispatcher process, local control units, and a data acquisition system. Each process is communicated to other processes and controlled by passing messages of commands and its status. A control flow and generalized command messages are defined following the software guideline for the SUBARU instruments. Functionality of each process is presented. Related off-line software for making multi-slit plates and for data analysis is briefly described.

A discussion of the tip tilt secondary control system developed for the NASA Infrared Telescope Facility is presented. Topics include: the use of existing IR drive electronics to drive an optical CCD at readout speeds approaching 300 Hz; outside suppliers for the secondary mirror and mechanical controls; the in house detector package creation; software system integration; and on-sky testing. The system strengths and weaknesses are discussed as they might apply to other projects. Tests of the system showed very good fast-guiding capability; revealed several telescope alignment problems; and allowed the IR instrument, NSFCAM, to produce science images with increased resolution. Work continues on improvement of ease of use, integration with the existing TCS control system, adaptation for use with other instruments and correction of outstanding problems with the secondary mirror support and surface quality.

An observation with Subaru Telescope is designed to be executed by the central scheduler process. Control commands are abstracted common to all observation instruments so that the observers are free from the consciousness of the difference between many instruments as much as possible. An abstraction command which is described in an observation procedure is expanded to a device dependent command script, and the script is dispatched to the telescope and instruments by referring to the instrument table. Device dependent commands are processed synchronously or asynchronously by checking the status against interlocks. The structure of the scheduler and the instrument table, the flow of commands such as an abstraction command, a device dependent command script, and a device dependent command, their examples and syntax are described.

Simple testing of concave telescope mirrors is discussed. Method uses a point light source in the infinity and a small lens in front of the focus of the mirror under test. Lens consists of two spherical surfaces and can be easily manufactured. Distance of the lens from the focus of the mirror under test is about a few centimeters, therefore the lens can be fastened to an interferometer for testing optics, to Foucualt or Ronchi tester, etc. It is necessary to use a flat mirror or collimator in an optical shop; in the space it is possible to use a light of stars. It is intended to use that method for controlling of the concave hyperbolic mirror of the Space Ritchey-Chretien telescope 1.7 m in diameter and f/2.74.

The United Kingdom Infra-Red Telescope (UKIRT) has recently undergone the first major upgrade program since its construction. One part of the upgrade program was an adaptive tip-tilt secondary mirror closed with a CCD system collectively called the fast guide system. The installation of the new secondary and associated systems was carried out in the first half of 1996. Initial testing of the fast guide system has shown great improvement in guide accuracy. The initial installation included a fixed integration time CCD. In the first part of 1997 an integration time controller based on computed guide star luminosity was implemented in the fast guide system. Also, a Kalman type estimator was installed in the image tracking loop based on a dynamic model and knowledge of the statistical properties of the guide star position error measurement as a function of computed guide star magnitude and CCD integration time. The new configuration was tested in terms of improved guide performance nd graceful degradation when tracking faint guide stars. This paper describes the modified fast guide system configuration and reports the results of performance tests.

This paper describes the design and current status of the mount control system (MCS) for the Gemini telescopes. The MCS is responsible for the interface between the telescope computer system (TCS) and the hardware systems that are used to move the telescope's two main axes (azimuth and elevation). In order to do this, the MCS must process encoder signals and use these to close a position servo involving multiple motors. The MCS also provides several ancillary functions. These are: the servos for the main axis cable wraps, the servos for the telescopes counterbalance units, an interface to the safety interlock system and inputs for various sensors that will be placed around the telescope structure.

Contact scientists for the Hubble Space Telescope examine observation products and report problems and shortcomings in the data. We have developed a database to hold this information and a series of World-Wide Web pages to facilitate reporting directly to this database. Over time, different characteristics of observations have been identified which are essentially signatures of each instrument that cannot be calibrated out. These have been classified as anomalies with which each contact scientist is very familiar. It is part of their task to identify these anomalies and flag them. Depending on the nature and severity of the anomaly, the principal investigator is contacted and the anomaly brought to their attention. To maintain a permanent record and to support others who retrieve data from the archive, we created a database for storing the contact scientists data quality assessments, and a WWW tool to facilitate the assessment process. The tool and database have been in routine use since early December 1996.

The gravity probe-B cryogenic star-tracking telescope provides the inertial pointing reference, as established by a distant guide star, with milli-arc-second resolution for the NASA/Stanford relatively gyroscope experiment. The star image of the f/27 Cassegrainian telescope is split onto two focal planes by a 50/50 intensity splitter, with each resultant image further divided by a roof prism reflector to generate the quadrant pointing information within few arc-seconds about the guide-star direction. Conventionally, the quadrant pointing information can be derived through the difference- and-sum algorithm. In this article, an alternative simple, yet robust algorithm is proposed and compared with the conventional one in the following aspects: (1) requirements on near perfect star-image division, (2) optimization in selecting null direction, (3) compensation of null-direction drift due to differential aging of photon detectors, (4) operational definitions of response sensitivity, linearity, and linear range of motion measurement, (5) robustness in system redundancy in terms of options in single-detector pointing per axis.

As a technical demonstration project for the NASA Advanced Communications Technology Satellite (ACTS), we have implemented remote observing on the 10-meter Keck II telescope on Mauna Kea in Hawaii from the California Institute of Technology campus in Pasadena. The data connection consists of ATM networks in Hawaii and California, running at OC-1 speeds (51 Mbit/sec) through optical fiber, and high data rate (HDR) satellite antennae at JPL in Pasadena and at the Tripler Army Medical Center in Honolulu. The ACTS network provides sufficient bandwidth to enable true remote observing, with a software environment identical to that used for on-site observing. In this paper, we demonstrate that while the satellite link introduces a number of difficulties and decreases overall reliability of the system, remote observing is not only feasible, but provides several important advantages over standard observing paradigms. Benefits include involving more members of observing teams while decreasing expenses, enhancing real-time data analysis of observations by persons not subject to altitude-related conditions, and providing facilities, expertise, and personnel not normally available at the observing site. Although the current bandwidth of the public Internet is insufficient for true remote observing, we nevertheless anticipate a growing role for remote observing techniques, particularly as high-speed terrestrial networking paradigms, such as ATM, become more commonly available.

The Galileo National Telescope (TNG) is a 3.6 meter Alt-Az telescope installed at the Astronomical Observatory of the Roque de Los Muchachos in La Palma, Canary Islands (Spain). The TNG motion control system, designed and realized by the Technology Working Group (TWG), is completely digital because of the versatility of this system topology. In a digital control system using an encoder as transducer means to have a digital feedback signal, therefore directly comparable with the reference without any conversion that is essential with other kinds of transducers. In the following the Galileo telescope (TNG) encoder system with its control electronics and the management software are described. It has been realized by a collaboration between the Heidenhain Company and the TWG. The TNG encoder system, at the state of the art, has one of the highest performances in the telescopes field, in terms of resolution, accuracy, readout time, reliability.

Recently, neural network models (NN), such as the multilayer perceptron (MLP), have emerged as important components for applications of adaptive control theories. Their intrinsic generalization capability, based on acquired knowledge, together with execution rapidity and correlation ability between input stimula, are basic attributes to consider MLP as an extremely powerful tool for on-line control of complex systems. By a control system point of view, not only accuracy and speed, but also, in some cases, a high level of adaptation capability is required in order to match all working phases of the whole system during its lifetime. This is particularly remarkable for a telescope control system. In fact, strong changes in terms of system speed and instantaneous position error tolerance are necessary. In this paper we introduce the idea of a new approach (NVSPI, neural variable structure PI) related to the implementation of a MLP network in an Alt-Az telescope control system to improve the PI adaptive capability in terms of flexibility and accuracy of the dynamic response range.

The Galileo National Telescope (TNG) is a 3.6-m Alt-Az telescope installed at the Astronomical Observatory of the Roque de Los Muchachos in La Palma, Canary Islands (Spain). The Galileo drive and control systems were designed and developed by the Technology Working Group (TWG) of the Capodimonte Astronomical Observatory, Naples (Italy). This paper presents a description of the dynamic model identification approach, and its use in the controller design.