A logical theory of localization (Q310092)

The authors consider a fundamental problem in cognitive robotics, namely the problem for a robot to identify its location and orientation to a reasonable certainty by means of available sensors and given a spatial characterization of its environment. The idea is to understand this localization problem as part of the situation calculus, a prominent first-order formalism for knowledge representation, in particular for reasoning about action and change. The central ingredient of the account given is an axiomatic basic action theory from which robot localization follows logically. This is illustrated by two examples of a robot in a two-dimensional grid, equipped with a moving action and distance sensor. Various properties of the basic action theory are shown to hold. Also, localization with multiple agents is treated.