Multi-fidelity approach for global trajectory optimization using GPU-based highly parallel architecture

Satoshi Ueda, Hideaki Ogawa

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)


A flexible and robust design space search method based on a systematic approach as well as effective optimization algorithms is required to accomplish challenging space missions. This paper presents a global trajectory optimization framework via a multi-fidelity approach that utilizes a graphics processing unit (GPU) for low-fidelity initial solution search and a central processing unit (CPU) to determine high-fidelity feasible solutions compliant with imposed constraints. The proposed framework consists of the following specific processes: (1) identifying a multitude of feasible trajectories with the aid of highly parallelized trajectory propagation using single-precision GPU cores; and then (2) determining accurate trajectories by means of gradient-based optimization incorporating double-precision propagation using CPU cores. The resultant trajectories have been verified in light of the primer vector theory that evaluates local optimality, thereby demonstrating the effectiveness of the proposed framework by robustly identifying optimum trajectories in the vast design space. Surrogate models have been constructed based on the solution database obtained in the course of optimization, enabling accurate prediction of trajectory attributes for given design parameters. Global sensitivity analysis has been performed using surrogate prediction to evaluate the influence of the decision variables on the trajectory parameters, which can be employed in multidisciplinary system design optimization.

Original languageEnglish
Article number106829
JournalAerospace Science and Technology
Publication statusPublished - Sept 2021

All Science Journal Classification (ASJC) codes

  • Aerospace Engineering


Dive into the research topics of 'Multi-fidelity approach for global trajectory optimization using GPU-based highly parallel architecture'. Together they form a unique fingerprint.

Cite this