Paper ID: 2202.09555

A Probabilistic Programming Idiom for Active Knowledge Search

Malte R. Damgaard, Rasmus Pedersen, Thomas Bak

In this paper, we derive and implement a probabilistic programming idiom for the problem of acquiring new knowledge about an environment. The idiom is implemented utilizing a modern probabilistic programming language. We demonstrate the utility of this idiom by implementing an algorithm for the specific problem of active mapping and robot exploration. Finally, we evaluate the functionality of the implementation through an extensive simulation study utilizing the HouseExpo dataset.

Submitted: Feb 19, 2022