Pseudospectral Optimal Control: Hidden Properties and Flight Results.

摘要

The main goal of this joint project between the Naval Postgraduate School and the University of California, Santa Cruz, was to analyze the mathematical proprieties of pseudospectral (PS) optimal control theory. The secondary goal of our project was to exploit the new properties to develop or enhance computationally efficient control algorithms. These goals were motivated by a need to generate real-time solutions to constrained nonlinear control problems that plays a central role in low-cost operations of high- performance space systems. In this project, we developed a general framework for PS optimal control theory based on discretization over an arbitrary (i.e. non-Gaussian) grid. Our analysis on dual consistency revealed some hidden properties of PS methods; and led to a new PS scheme utilizing primal-only conditions. These new developments enhance the algorithmic efficiency that is crucial for real-time applications. During this research period, we also got an opportunity to apply our ideas to the time-optimal reorientation of a NASA spacecraft in orbit; and performed a historic command and control of the Transition Region and Coronal Explorer (TRACE) space telescope.