On-Board Optimal Feedback Controller Generation for Hypersonic Re-Target Scenarios

This paper presents a framework for rapid on board generation of hypersonic trajectories. We specifically focus on the hypersonic re-target scenario, which occurs when a vehicle is launched with a nominal target and corresponding optimal trajectory, but receives a new terminal target after the vehicle has already flown part of the nominal trajectory. Qualitatively speaking, we present a novel blend of data-driven learning approaches with indirect optimal control techniques involving banks of open-loop trajectories. The learning is accomplished using sparse approximation techniques resulting in a numerically parsimonious surface fit that is well-suited for on board computations. As part of the re-targeting mission, the vehicle uses this surface fit to generate optimal feedback controllers in real-time. We demonstrate the application of our proposed framework for planar hypersonic missions with re-targeting.