Efficient scheduling of astronomical surveys is a challenge with an increasing level of complexity as the observation strategies are becoming more sophisticated and operational costs are higher. In general, any kind of astronomical survey requires the execution of a huge number of observations fulfilling several constraints. The fulfillment and optimization of these constraints is a key factor for obtaining an efficient schedule with an adequate exploitation of the resources and with a high scientific return. In this contribution, we present the framework STARS (Scheduling Telescopes as Autonomous Robotic Systems) that computes optimal schedules for a variety of space- and ground-based infrastructures and scientific exploitation plans. STARS provides methods, tools and libraries for the definition of surveys (e.g., objects to observe, features of the objects, observation constraints), the definition of the observatories (e.g., location, number of telescopes, type of telescopes, sub-array configurations), the usage of astronomical calculations (e.g., object coordinates, object elevation, Sun and Moon position, Moon phase), and the application of schedulers (e.g., long-term, short-term) based on Genetic Algorithms (GAs) and astronomy-based heuristics.
In STARS, two main types of schedulers are defined: long-term and short-term. The long-term scheduler is focused on scheduling object observations with a time scope ranging from one night to several months or years. It considers the observation constraints (hard-constraints) that can be predicted beforehand, and it optimizes some objectives (soft constraints) by using GAs. The execution of the long-term scheduler can be time-expensive, but it is not time-critical because it can be run before the start of the telescope operation, so it can be used as a standalone scheduling tool. On the other hand, the short-term scheduler computes in real-time the next observation (or scheduling block) to be executed by optimizing some soft constraints, fulfilling all the hard constraints and by considering all the observations previously executed. The short-term scheduler is time-critical and reacts in less than a second to the changing conditions (weather, errors, delays, targets of opportunity). It uses astronomy-based heuristics to repair the schedule obtained by the long-term scheduler, in order to keep the long-term perspective while avoiding intensive calculations.
STARS has been successfully applied in several ground and space-based observatories. It is used to operate the CARMENES instrument (Calar Alto, carmenes.caha.es) and the Joan Oró robotic Telescope (www.oadm.cat). It is used to prototype the mission planning tool for the ARIEL M4-ESA candidate mission, and in prototypes for large ground-based installations, such i.e. the Cherenkov Telescope Array (CTA). Finally, STARS is also being extended to cover multi-observatory coordinated scheduling purposes, under the framework of the EU-H2020 ASTERICS project, and in order to promote multi-messenger science. The coordination of large observatories in the northern and southern hemispheres are used as test cases to evaluate the performance of such an innovative scheduling solution. In this sense, simultaneous observations or minimal time gap between observations are promoted resulting in a challenging and complex optimization problem that will open a new era for the optimal operation of large astrophysical infrastructures.