The aim is to investigate the use of Deep Neural Networks (DNNs) in combination with Dynamical Systems Theory (DST) to develop an intelligent initial guess generator capable of speeding up the optimisation process of low-energy interplanetary trajectories in the CR3BP model, while achieving more efficient (reduced ΔVcost) and accurate trajectories. In particular, the focus is on enhancing low-thrust end-to-end transfer mission designs between moons in the framework of the Jovian System by exploiting the low-energy structures associated with resonant flybys.