General and efficient methods for queries and optimization in uncertain co-design

Develop general and efficient computational methods for queries and optimization in uncertain co-design, where uncertainty is modeled as probability distributions over monotone design problems and as uncertain parameterized design problems represented by Markov kernels, and where composition operations are lifted to these uncertain structures.

Background

The paper extends the monotone co-design framework to incorporate interval, distributional, and parameterized uncertainty and shows how standard co-design composition operations lift to these semantics. In this setting, uncertain parameterized design problems are modeled as Markov kernels, enabling conditional distributions over design problems given parameters.

This enriched semantics introduces new challenges for solving co-design queries (fix functionalities minimize resources; fix resources maximize functionalities) and for performing optimization over design choices when uncertainty is present, particularly when choices must be made before observing parameter realizations. The authors explicitly note that the development of general and efficient methods to address such queries and optimization tasks under uncertainty remains open.