SOFIA observatory automated scheduling after 5 years of operations

This paper describes a new framework for scheduling that has been developed for the NASA Stratospheric Observatory for Infrared Astronomy (SOFIA). Key to successful and cost-efficient operations of the SOFIA airborne observatory is the optimized scheduling of operational activities. These include instrument, observation and maintenance schedules, as well as Southern Hemisphere deployments. The most distinctive aspect of the SOFIA flight scheduling problem is the interdependency of the targets than can be observed in a same flight, which makes automated scheduling techniques available for ground-based and space-based telescopes unsuitable. SOFIA began early science operations in 2011 and is currently completing its fourth annual cycle of operations, which consists of about 550 hours of observer time carried out during ~100 science flights. Although early conceptual studies on the SOFIA scheduling problem were previously conducted, flights still had to be manually created when operations started. Here, we introduce the new automated scheduling system based on a tree search algorithm that is used to generate long-term and short-term operational schedules. We provide a formulation of the SOFIA scheduling problem, as defined after 5 years of operations, including all constraints that a valid schedule must satisfy. We list the flight operational tasks that must be efficiently simulated while building the global search tree. We discuss the foundations of the scheduler and describe the constraint representation, algorithm and heuristics that guide the search. Finally, we report on the integration of the automated system in mission operations and its current and future expected performance.