Improving science yield for NASA Swift with automated planning technologies

TL;DR

This paper formalizes the scheduling challenge of the Swift mission as a dynamic fuzzy constraint satisfaction problem and explores advanced optimization, machine learning, and neural network methods to automate and improve scheduling quality.

Contribution

It introduces a formal model for the Swift scheduling problem and investigates novel algorithmic and machine learning approaches to surpass human scheduling performance.

Findings

Formalized Swift scheduling as a DF-CSP

Explored classical and machine learning optimization techniques

Discussed potential for increased scientific yield

Abstract

The Swift Gamma-Ray Burst Explorer is a uniquely capable mission, with three on-board instruments and rapid slewing capabilities. It serves as a fast-response satellite observatory for everything from gravitational-wave counterpart searches to cometary science. Swift averages 125 different observations per day, and is consistently over-subscribed, responding to about one-hundred Target of Oportunity (ToO) requests per month from the general astrophysics community, as well as co-pointing and follow-up agreements with many other observatories. Since launch in 2004, the demands put on the spacecraft have grown consistently in terms of number and type of targets as well as schedule complexity. To facilitate this growth, various scheduling tools and helper technologies have been built by the Swift team to continue improving the scientific yield of the Swift mission. However, these tools have been used only to assist humans in exploring the local pareto surface and for fixing constraint violations. Because of the computational complexity of the scheduling task, no automation tool has been able to produce a plan of equal or higher quality than that produced by a well-trained human, given the necessary time constraints. In this proceeding we formalize the Swift Scheduling Problem as a dynamic fuzzy Constraint Satisfaction Problem (DF-CSP) and explore the global solution space. We detail here several approaches towards achieving the goal of surpassing human quality schedules using classical optimization and algorithmic techniques, as well as machine learning and recurrent neural network (RNN) methods. We then briefly discuss the increased scientific yield and benefit to the wider astrophysics community that would result from the further development and adoption of these technologies.