GALCIT Colloquium

Since the dawn of modern computing, hypersonic mission design has been largely accomplished using direct optimization methods. While these methods have enabled the practical optimization of hypersonic trajectories, useful analytic information about the design problem is lost. This analytic information provided by indirect optimization methods has been viewed as unavailable due to the limitations associated with the technique. Research by Dr. Grant and his students at Purdue has illustrated that the historical limitations associated with indirect optimization methods can be overcome to perform rapid, high quality mission design of hypersonic systems in a largely automated manner. Example mission design scenarios of various hypothetical systems illustrate the fidelity and quality of the approach.

The natural parallelization of the indirect methodology has also enabled execution on highly parallel graphics processing units by removing the large coupled calculations associated with direct optimization methods. Algorithmic optimizations have resulted in speedups compared to traditional CPU-based algorithms for single vehicle hypersonic mission design. This is particularly noteworthy due to the relatively small problem size associated with single vehicle trajectory optimization. The indirect methodology is expected to also address future, large-scale mission design problems (e.g., to design emergent swarm behavior for operation in highly dynamic environments) without a substantial reduction in solution quality or speed.