How Albot1 Computes its Topological-metric Map

Abstract

Recent research study has proposed a computational theory of perceptual mapping that shows how a useful inexact global perceptual map is computed without correcting sensor errors. An unexpected consequence of the theory is that the map computed is never closed, expanding ad infinitum. This creates an interesting transient perceptual map but then an important question arises: how does one compute an enduring map from it? In this paper, we address this question by showing how Albot1 computes a topological-metric map from its perceptual map. Importantly, we show that the map computed is not based upon partitioning the physical environment into places of a fixed size and that these places, when combined using a single frame of reference, do not produce a precise global map. We show how such a map is used during navigation to provide expectations, say, to where exits in a current place can be found.