"Tracking Regions using Conceptual Neighbourhoods"

 

David Randell and Mark Witkowski

Dept. of Electrical and Electronic Engineering
Imperial College of Science, Technology and Medicine
Exhibition Road
London SW7 2BT
United Kingdom
{d.randell, m.witkowski}@imperial.ac.uk

 

Abstract

The use of conceptual neighbourhood diagrams (CNDs) is now fairly well established in AI literature. In AI applications, and in particular Qualitative Spatial Reasoning (QSR), the nodes of CNDs are typically populated by individual relations defined on regions or intervals, and the edges connecting adjacent nodes as continuous transformations between them. While CNDs have been used to measure the similarity between relations and their relata, region-identity is often assumed rather than explicitly determined by the representational theory used.  In this paper we explore QSR theories and relation-based measures of similarity for region-identity in a dynamic setting.  Ambiguity in potential identity mappings arising from using a weak qualitative similarity measure is reduced by combining conceptually related but distinct QSR theories, each supporting a CND. Results from a simulation program that implements the theory are discussed.