Christopher Valicka, Sandia National Laboratories, Deanna Garcia, Sandia National Laboratories, Andrea Staid, Sandia National Laboratories, Jean-Paul Watson, Sandia National Laboratories, Mark D. Rintoul, Sandia National Laboratories, Gabriel Hackebeil, University of Michigan, Lewis Ntaimo, Texas A&M University,
Keywords: Scheduling, SSA, stochastic optimization
Abstract:
Advances in space technologies continue to reduce the cost of placing satellites in orbit. With more entities operating space vehicles, the number of orbiting vehicles and debris has reached unprecedented levels and the number continues to grow. Sensor operators responsible for maintaining the space catalog and providing space situational awareness face an increasingly complex and demanding scheduling requirements. Despite these trends, a lack of advanced tools continues to prevent sensor planners and operators from fully utilizing space surveillance resources. One key challenge involves optimally selecting sensors from a network of varying capabilities for missions with differing requirements. Another open challenge, the primary focus of our work, is building robust schedules that effectively plan for uncertainties associated with weather, ad hoc collections, and other target uncertainties. Existing tools and techniques are not amenable to rigorous analysis of schedule optimality and do not adequately address the presented challenges. Building on prior research, we have developed stochastic mixed-integer linear optimization models to address uncertainty due to weather’s effect on collection quality. By making use of the open source Pyomo optimization modeling software, we have posed and solved sensor network scheduling models addressing both forms of uncertainty. We present herein models that allow for concurrent scheduling of collections with the same sensor configuration and for proactively scheduling against uncertain ad hoc collections. The suitability of stochastic mixed-integer linear optimization for building sensor network schedules under different run-time constraints will be discussed.
Date of Conference: September 20-23, 2016
Track: Poster