?Icy moons of the outer Solar System, such as Europa and Enceladus, are among the most compelling targets
in the search for habitable environments beyond Earth, due to strong evidence of frozen crusts and subsurface
oceans. In particular, Europa has emerged as a primary scientific focus, motivating dedicated exploration
campaigns aimed at characterizing its ice shell and its ocean, such as NASA’s Europa Clipper mission.
Achieving these scientific objectives relies on repeated close flybys and complex transfers.

Deep-space missions operating in highly perturbed environments require trajectory design methodologies
that explicitly account for uncertainties such as navigation, dynamic modeling and maneuver execution er-
rors. This need is significant in the Jovian system which features locally chaotic nonlinear dynamics and
involves frequent Jupiter’s moons flybys that demand precise targeting and guarantees of safety against
unexpected loss of spacecraft maneuverability.
Traditional design practice relies on deterministic reference trajectories augmented a posteriori with heuristic
safety margins and extensive navigation analysis. However, there has been a growing number of studies on
directly incorporating these statistical effects into mission design and coupling the trajectory optimization
with Navigation processes within the design process.

Motivated by recent advances in stochastic optimal control and convex programming for space trajectories,
this work presents a sequential covariance steering approach to design an optimal trajectory and its correc-
tion policies that can probabilistically guarantee safety constraints under the assumed error models. In this paper the algorithm is demonstrated on several transfers in the Jupiter-Europa Circular Restricted Three Body Problem.

Two distinct convex formulations for covariance propagation are implemented and compared: (i) a block
Cholesky method, which parameterizes the covariance evolution through a large block lower-triangular ma-
trix, and (ii) a full covariance method, which optimizes the covariance matrices with an elegant lossless
convexification of originally non-convex covariance propagation equations by using the Schur complement
and the Karush-Kuhn-Tucker (KKT) conditions. Although both approaches have been proposed in prior
work, they have not been systematically compared within the same context.
The optimization incorporates B-plane geometric targeting constraints at each flyby, imposed on both the
mean arrival state and its covariance, ensuring that the geometry requirements are met. In addition, a safety
chance constraint is enforced to guarantee a minimum flyby altitude even in the event of total loss of control
from any point onward along the trajectory. This results in a trajectory that remains safe under a wide
range of failure scenarios without requiring externally tuned margins.
The framework is applied to representative Jupiter-centric transfers from the Europa Clipper science phase,
including resonant and non-resonant arcs. For each segment, the sequential convex programming (SCP) loop computes both the nominal control history and a linear feedback policy required for the covariance steering. The objective is to minimize the ?V99 or an appropriate upper bound of it while satisfying all probabilistic constraints.
To further decrease the number of feedback maneuvers applied along the trajectory, an ?1,p-regularized
sparsification algorithm is introduced inspired by recent developments in hands-off covariance steering. The ?regularization penalizes the activation of the feedback corrections in an iteratively reweighted way, effectively
promoting solutions in which a feedback maneuver is engaged only when strictly necessary to control the
growth of uncertainty. For typical science-phase flybys, the approach achieves targeting accuracies compat-
ible with mission requirements while maintaining compliance with safety constraints at the 99 % level.
After obtaining the optimal trajectory and feedback policy from the SCP
procedure, a nonlinear Monte Carlo (MC) validation is performed to assess the fidelity of the linearized
assumptions used in the covariance steering formulation. In this step, the full nonlinear dynamics, the
nonlinear measurement model, and all uncertainty sources are reintroduced and propagated throughout the
trajectory.

Overall, the study demonstrates that convex stochastic control techniques provide a rigorous and computa-
tionally efficient methodology for designing robust multi-flyby transfers in strongly perturbed environments.
For missions like Europa Clipper, where safety and science return critically depend on precise flyby geometry
under uncertainty, these methods enable integrated design of the nominal path, uncertainty-shaping feed-
back, and safety guarantees, reducing reliance on iterative navigation analysis. The framework also holds
promise for future applications to Jovian and Saturnian moon tours and for integration into autonomous
onboard guidance architectures capable of real-time robust decision-making.