Representing and Reasoning with Logical and Probabilistic Knowledge on Robots

This paper describes an architecture that enables robots to represent and reason with logic-based and probabilistic descriptions of uncertainty and domain
knowledge. An action language is used to describe tightly-coupled transition diagrams of the domain at two different resolutions. For any given goal, reasoning with commonsense knowledge at coarse-resolution provides a plan of abstract actions. To implement each such abstract action, the
relevant part of the fine-resolution transition diagram is identified, and used with a probabilistic representation of the uncertainty in sensing and actuation, to obtain a sequence of concrete actions. The outcomes of executing these concrete actions are used for subsequent coarse-resolution reasoning. We illustrate the architecture’s capabilities using examples of a mobile robot finding and moving objects in an indoor domain.