Spike: AI scheduling for NASA's Hubble Space Telescope

Abstract

The Spike scheduling system, developed for scheduling astronomical observations for NASA's Hubble Space Telescope (HST), is described. Spike is an activity-based AI scheduler which incorporates innovative approaches to constraint representation and reasoning and scheduling search. Although developed for space telescope scheduling, the Spike system was designed around a general scheduling-constraint framework that can be applied to other domains. Techniques for machine reasoning about scheduling constraints and goals, especially in cases where uncertainty is an important scheduling consideration and where resolving conflicts between competing preferences is essential, have been developed. These techniques have been used in the Spike system. Graphical displays of activities, constraints, and schedules are an important feature of the system. High-level scheduling strategies using both rule-based and neural network approaches have been developed. While the specific constraints implemented are those most relevant to the HST, the Spike framework is more general and could handle other kinds of scheduling problems. >