Reproduce “Asteroid Flyby Cycler Trajectory Design Using Deep Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Naoya Ozaki, Kanta Yanagida, Takuya Chikazawa, Nishanth Pushparaj, Naoya Takeishi, Ryuki Hyodo
Ozaki et al. from JAXA and the University of Tokyo developed a Deep Neural Network (DNN)-based approach for designing asteroid flyby cycler trajectories, efficiently generating training data through "pseudo-asteroids" and reducing computation time for identifying optimal sequences from an estimated 7 days to approximately 10 hours.
Key insights
- Generating pseudo-asteroids based on KKT conditions offers a scalable and computationally inexpensive method to overcome the data scarcity for DNN training in astrodynamics.
- Integrating astrodynamics knowledge, such as free-return trajectory parameters and Lambert's problem results, into DNN inputs enhances prediction accuracy and model reusability across diverse mission parameters.
- DNNs can serve as highly accurate and efficient surrogate models for complex trajectory optimization problems, significantly accelerating global search algorithms compared to traditional direct optimization or simpler heuristics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.