Reinforcement Learning for Sequential Low-Thrust Orbit Raising Problem

In this paper, we consider the electric orbit-raising problem posed as a sequence of multiple trajectory optimization sub-problems. Each sub-problem aims to move the spacecraft closest to the geosynchronous equatorial orbit (GEO) by minimizing a convex combination of suitably selected objectives. Recognizing that the weights in the objective function influence the transfer time of the computed low-thrust orbit-raising trajectory, we propose a mathematical formulation in the framework of reinforcement learning in order to enable adaptive modification of the weights during the transfer. Artificial neural networks are constructed and trained on orbit-raising scenarios in order to compute the reward functions associated with reinforcement learning. Numerical simulations for planar and non-planar transfer scenarios are provided to demonstrate the proposed methodology.