A Fast Solution to Impulse Controllable Set of Direct Abort Orbits From Low Lunar Orbit Based on Deep Learning

To ensure the safety of astronauts during crewed lunar exploration missions, it is crucial to develop a fast solution to the controllable set (CS) of direct abort orbits from Low Lunar Orbit (LLO). This article proposes a learning-based method for rapid calculation of the CS. First, a definition and mathematical formulation of the CS are developed. Then, a semi-analytical calculation method, incorporating the pseudo-perilune parameter set and a two-segment double two-body model, is proposed to calculate the CS databases under three different cases. Next, a Fully Connected Neural Networks (FCNN)-based solution for the CS is proposed, and its validation is performed through numerical simulations. The results demonstrate that the proposed learning-based method significantly improves calculation efficiency compared to traditional methods, while maintaining equivalent accuracy. Finally, parameter characteristic analysis reveals the influence of constraint conditions on the topological and numerical characteristics of the CS through a large number of numerical simulations. These findings provide a rapid solution and analysis tool for determining direct abort orbits in future crewed lunar exploration missions, offering theoretical foundations and engineering guidelines for optimizing emergency escape strategies from LLO.