The Zwicky transient facility robotic observing system

The transient universe is fast becoming one of the most important research areas in astronomy. Finding objects that change, either quickly or periodically, has opened up new understanding of the cosmos around us, and brought up new questions that require further investigation. The Zwicky Transient Facility (ZTF) has been developed to observe as much of the sky as possible at a rapid rate, in order to expand the regime of time domain measurement to shorter intervals and detect changes in the sky more quickly. ZTF is a fully automated system, composed of the Samuel Oschin 48-inch (1.2m) telescope at Palomar Observatory (P48), the mosaic camera constructed by Caltech, a filter exchange system, associated sensors and electrical systems, and the Robotic Observing Software (ROS) that controls the operation of the entire system. P48 is a 70 year old telescope that has been upgraded with new hardware, electronics, and a modern telescope control system to allow it to move quickly and accurately across the sky under robotic control. The ZTF mosaic camera is a custom system composed of 16 6Kx6K pixel CCDs, creating a mosaic camera with over 576 million pixels that can image 47 square degrees down to a magnitude of 20.5 in a 30 second exposure. The filter exchange system uses a Kuka robotic arm to grab the 400x450mm filters out of a storage closet and place them onto the front of the mosaic camera, where they are held in place by electromagnets and locking pins. A full sensor system monitors the health of the camera dewar and environment of the observatory; a separate weather station monitors the outside environment. Other subsystems control the motion of the Hexapod that the mosaic camera is mounted on, the top end shutter, and remote switching of power, Managing all of these subsystems is ROS, which is the automated control software that runs ZTF observations. ROS is based on the Robo-AO control system, with improved automation procedures and expanded capabilities to handle the operations required for ZTF. ROS consists of 31 separate software daemons spread across 5 computer systems (4 to control the mosaic camera, 1 for robotic operation); the robotic control daemon is able to manage all daemons, as well as start and stop their operation as necessary. Watchdog daemons intervene in case of robotic system problems, and each daemon has an internal watchdog that can fix or kill the daemon in case of difficulties; if a daemon dies the robotic system automatically restarts it. ROS controls the start of observations and morning shut down, handles weather monitoring and safely stopping in case of bad weather, and responds to problems in the observing sequence by fixing them or stopping operations and sending a message for help. All calibration measurements are done automatically at the beginning of the night; if the calibrations are interrupted they are completed after observations finish in the morning. A queue system determines the observation priority and revises the order of observations dynamically to optimize observational efficiency. ROS is able to operate with less than 15 second overhead between each standard ZTF observation (with a 7.5 degree slew); this is achieved by reading out thee mosaic camera during telescope slew, then transferring and writing FITS data files during the next exposure. FITS headers are kept synchronized through daemons that gather all relevant FITS header information and distribute that to the camera computers. ROS is able to produce more than 80 mosaic science images per hour in standard survey mode; each mosaic is a total of 380MB compressed, so the system produces more than 30GB of data on disk per hour that have to be transferred off the mountain. A new data transfer system synchronizes the compressed FITS data files to the data analysis servers in Pasadena, CA in parallel with the observing system; images are in place for the data analysis pipelines in less than a minute after the ZTF shutter closes. This presentation will discuss the development and execution of the ZTF observing software, as well as analyze the observational behavior and efficiency of the system during the first few months of on-sky science observations