Parallel Nonlinear Optimization for Astrodynamic Navigation

CU Aerospace proposes the development of a new parallel nonlinear program (NLP) solver software package. NLPs allow the solution of complex optimization problems, and there exist today a number of capable NLP solvers typically used within the astrodynamics community. Unfortunately, none of them take into account the capabilities of a high performance computer with thousands of processors, nor use the recent advances in graphics processing unit (GPU) calculation. Many NASA optimal trajectory packages (OTIS, MALTO, EMTG) have already streamlined and optimized their own code to run as fast as possible, and it is now the serial-execution nature of the existing packages that represents the largest bottleneck. The new parallel NLP solver to be developed by CU Aerospace in partnership with the University of Illinois will represent a novel, ground-up redesign of this kind of solver. It will be built to be transparently usable by the existing optimal trajectory solvers used at NASA; however, it will also take advantage of high performance computing and/or GPU processing to reduce the run time by orders of magnitude. This has strong implications for NASA's mission design groups, as the time to solution for the trajectory teams will be significantly improved. Further, an improved optimization tool will have applications across all fields at NASA which user optimization, from engineering design to cost analysis. CU Aerospace is uniquely positioned to develop this new parallel NLP solver, with a team of experts, and the nearby proximity to the high performance computing resources at the National Center for Super Computing.