Reasoning About and Scheduling Linked HST Observations with SPIKE

Scientific observations executed by the Hubble Space Telescope (HST) are subject to a number of complex and interacting timing and orientation constraints. We classify these constraints as " absolute " and " relative. " The former apply to only one observation, whereas the latter apply to two or more linked observations or " visits. " Examples of absolute constraints are " Schedule Visit A BETWEEN Day 10 and Day 20 " or " Schedule Visit A at an ORIENT in the range 30 to 40 degrees. " Relative constraints express either how the visits should be linked in time (for example " Schedule Visit2 AFTER Visit1 by 10 days " or linked in orientation (for example " Schedule Visit2 ORIENT FROM Visit1 by 20 degrees "). This paper will focus on the implementation of relative constraints , how they are combined with absolute constraints, and the problems encountered and solved in combining relative constraints along the orthogonal dimensions of time and orientation. We also briefly present the SPIKE Plan Window Scheduler (Giuliano 1997), which creates a Long Range Plan by assigning " plan windows " where the visits might feasibly be scheduled, and which optimizes the placement of windows to satisfy a number of criteria. Since the assigned plan windows are much larger (nominally eight weeks) than the actual times needed to schedule the visits, we confront the problem of " link set flexibility. " This implementation allows a great deal of latitude for a short term scheduler to later assign actual spacecraft execution times to visits. We also describe an enhancement of the Plan Window Scheduler to schedule orientation angles along with plan windows for orient linked visits, while confronting computational constraints. The SPIKE (Johnston and Miller 1994) software system is used both to determine where Hubble Space Telescope observations are schedulable and to craft a long range observing plan that is flexible while maximizing telescope usage. Scheduling in this large, complex domain is further complicated by dependencies between individual observations , or visits, in an observing program. Approximately 39% of 2,444 HST visits for Cycle 7 are linked by timing and/or orientation constraints, and ensuring that these observations are scheduled correctly and planned optimally is a major operational concern for the Hubble ground system. In this paper we describe problems that have arisen in reasoning correctly about linked observations and our solutions to these problems. We go on …