Neural Network-Based Approximation Model for Perturbed Orbit Rendezvous

An approximation of orbit rendezvous is usually used in the global optimization multi-target missions, which can greatly affect efficiency process. A fast neural network-based surrogate model proposed to approximate optimal velocity increment perturbed low Earth orbits. According a dynamic analysis, initial and target orbits together with flight time are transformed into nine-dimensional normalized vector that as input layer network. existing method introduced quickly generate training data. In simulations, different numbers nodes hidden layers tested choose best parameters. The network demonstrates high precision compared previous methods models. mean relative error less than 1%. Finally, case an mission prove potential application model.