太陽系の遠方天体（木星以遠の惑星やメインベルト以遠の小惑星など）の探査においては，燃料消費を抑えつつ目標天体へ到達するため，複数の惑星重力を利用して速度変更を行う「多重重力アシスト（Multiple Gravity Assist, MGA）」が不可欠である．しかし，MGA軌道設計は惑星遭遇順序，遭遇時刻，飛行時間など多くの設計変数を含む複雑な組合せ最適化問題であり，150万を超える膨大な数の小惑星候補に対して個別に軌道の最適化を行うことは計算コストの観点から困難である．
本研究では，従来よりエネルギー的な観点でのMGAシーケンスの絞り込みに用いられてきた“Tisserand Graph”を基盤としつつ，動的なグラフを構築してグラフ構造全体を学習することで，任意の天体に到達するためのMGA軌道設計の高速化を目指す．

In the exploration of distant bodies in the solar system (planets beyond Jupiter, asteroids beyond the main belt, and so on), the Multiple Gravity Assist (MGA), which changes velocity by using the gravity of several planets, is indispensable for reaching the target body while keeping propellant consumption low. However, MGA trajectory design is a complex combinatorial optimization problem involving many design variables such as the planetary encounter sequence, encounter times, and flight times, and individually optimizing the trajectory for the enormous number of more than 1.5 million asteroid candidates is difficult from the standpoint of computational cost.
In this research, while building upon the "Tisserand Graph"—which has conventionally been used to narrow down MGA sequences from an energetic standpoint—we construct a dynamic graph and learn the entire graph structure, thereby aiming to accelerate the MGA trajectory design needed to reach any target body.