I. Introduction
In recent years, modern robots have been breaking the barriers of laboratories into the real world. In real scenarios, accounting for different sources of uncertainty is essential, both in state inference and decision making, such that truly autonomous, reliable and robust performance can be attained. Moreover, such settings often involve reasoning about a belief over a high-dimensional state. Relevant problems include SLAM and autonomous navigation in unknown environments. Causes for the uncertainty in that context can be, for instance, noisy measurements, failed locomotion attempts (e.g. wheels slippage) and inherit risk taking.