What Observations Really Tell Us

摘要

When agents like mobile robots make observations while carrying out a course of actions, a formalization of the observed information is needed in order to reason about it. When doing so in the situation calculus, a seemingly straightforward approach turns out to be inappropriate since it leads to unintended results and has an unfortunate sensitivity with respect to different forms of successor state axioms. In this paper we suggest how to properly encode observed information in order to avoid both of these problems.