Multi-objective robust trajectory optimization of multi asteroid fly-by under epistemic uncertainty

Methods are proposed and compared to generate robust optimal trajectories subject to epistemic uncertainty, meaning uncertainties that derive from a lack of knowledge on system's and launcher's parameters.
This type of uncertainty is typical of the early stage of the design process when multiple options need to be evaluated and only a partial knowledge of each of them is available.
The uncertainty is modelled using probability boxes (p-boxes) and lower expectation. The p-box is a family of distributions that is known to contain the real probability distribution, and the lower expectation is the minimum expectation that can be obtained with distributions within that family. We test multiple methods for efficiently estimating this quantity.
These lower expectations are optimised using a Multi-Objective solver MACS (Multi Agent Collaborative Search), and with surrogate models to speed-up the optimization.
Furthermore, novel dimensionality reduction methods are employed, based on control mapping, as well as a method, threshold mapping, that improves the quality of the optimization by focusing the search on target sets that produce non-trivial values of the lower expectation.