MULTIPLE IMPULSE CISLUNAR TRANSFER VIA MULTI-FIDELITY APPROACH USING GPU-BASED SUPER PARALLELIZATION

This paper presents a global trajectory optimization framework via a multi-fidelity approach that makes combined use of a graphics processing unit (GPU) for a low-fidelity initial solution search and a central processing unit (CPU) for high-fidelity optimization to determine solutions that satisfy constraints including Lawden’s conditions in light of primer vector theory. The present study aims to identify optimum solutions robustly and flexibly by introducing multiple impulses that can further reduce the total velocity increments compared with two-impulse transfer. The proposed framework is examined with a mission scenario considering cislunar transfer from a near-rectilinear halo orbit (NRHO) to a low lunar orbit (LLO).