Multi-objective Design of Direct Transfers between Libration Point Orbits via Deep Learning Method

This paper presents a nested deep learning-based multi-objective optimization framework for designing direct transfer trajectories between libration point orbits in the Circular Restricted Three-Body Problem (CRTBP). Traditional single-objective optimization methods typically focus on either fuel efficiency or transfer duration, often overlooking the inherent tradeoffs essential for practical mission design. By integrating deep neural networks (DNNs) with metaheuristic algorithms, the proposed approach simultaneously optimizes fuel consumption, transfer duration, and trajectory stability while ensuring direct transfers without intermediate orbital corrections. Simulation results in the Earth-Moon system demonstrate that, compared to conventional numerical solvers, the proposed method reduces computational costs by 60% while successfully generating Pareto-optimal solution sets. This provides mission designers with flexible multi-objective trade-off strategies, making the framework well-suited for deep space exploration trajectory planning.