Transformer-Informed Trajectory Optimization for Relative Motion in Cislunar Orbits

This paper presents a Transformer-based method to generate near-optimal initial guesses for trajectory optimization in cislunar space. The approach addresses the challenge of severely nonlinear and chaotic dynamics in cislunar orbits compared to Earth orbit. By providing warm-starting solutions, the method accelerates convergence of optimization protocols for autonomous spacecraft guidance, rendezvous, proximity operations, and docking applications.