Designing robust, efficient, and autonomous spacecraft trajectories under uncertainty is increasingly critical for modern space missions, especially those involving small spacecraft and deep-space operations where onboard autonomy and limited computational resources are key constraints. Traditionally, trajectory design is performed deterministically, followed by expensive Monte Carlo analyses to assess robustness, a time-consuming approach that delays mission development and might compromise optimality. This thesis introduces a framework for trajectory design and control that integrates uncertainties from the preliminary design, paving the way for faster, more resilient, and autonomous spaceflight operations.

The work first presents a novel uncertainty propagation technique capable of modeling non-Gaussian distributions through nonlinear dynamics with high accuracy and significantly reduced computational cost. By leveraging high-order Taylor expansions and moment generating functions, the method enables analytical propagation of statistical moments without reliance on sampling. This is applied across various mission scenarios, including the European Space Agency’s Hera mission to Didymos-Dimorphos.

Building on this foundation, the thesis introduces a stochastic continuation framework that generalizes classical orbit continuation methods to account for uncertainty. This framework enables the systematic design of trajectories that are inherently robust by embedding uncertainties into the continuation process itself. The methodology is demonstrated in both planar and three-dimensional settings of the Circular Restricted Three-Body Problem and in space missions-relevant scenarios, such as the European Space Agency’s Hera mission, to showcase its use for identifying resilient trajectory families under uncertainty.

Finally, the thesis explores the use of Neural Ordinary Differential Equations for uncertainty-aware control, after introducing their use in space trajectory design. This approach augments deterministic controllers with lightweight neural corrections trained to reduce terminal state dispersion, offering a new way to make deterministic neural control policy resilient to uncertainties.

Together, these contributions establish a framework for uncertainty-aware trajectory design and control, with broad applicability to autonomous, resource-constrained missions of the future.