Optimal Observation of Satellites Using Combined Measurements from Many Networked Observers Cooperatively Controlled by Human and Autonomous Means

This paper develops methods for regularly characterizing satellites at high resolution and low recurring cost using satellites whose combined orbits collectively have excellent views of space objects. Humans and autonomous spacecraft efiectively collaborate by allowing humans to modify and direct both overall objectives and short term goals while using autonomous planning techniques to manage complex, yet well-deflned and predictable orbital events and sensor allocations. This allows humans to provide input where they are strong, such as understanding high level goals and reasoning amidst uncertainty. Simultaneously, it allows autonomous optimization and planning techniques to operate in domains where they are strong, in this case when many spacecraft must be jointly controlled over space and time to achieve a clearly deflned overall flnal result. Characterization of satellites in geosynchronous orbit is especially important, as they are too far away for high resolution observation from Earth. This paper formulates observation of multiple satellites by a set of Observer satellites flrst as a binary integer programming problem. Next, methods are found to relax the formulation to become a convex linear program. Thus global optimality, high speed, and convergence to the optimum are assured. Real-time human input is allowed by modiflcation of the optimization weights. For instance, if a human decides that a particular object requires extra observation immediately, then this paper shows how this short term goal can be achieved by weight modiflcation. Similarly, if a human operator desires that an object be observed more persistently and accurately than other objects, then this paper develops weighting methods to achieve this long term goal. Simulation results conflrm the viability of the technique on a variety of orbits, including both low Earth and geosynchronous cases. In addition, the simulations indicate that the binary approximation of the convex relaxation achieves nearly optimal performance. The method is fully general, encompassing 3-D and elliptical orbits.