A rule-based genetic algorithm for mapping route descriptions towards map representations

Abstract

When maps are not available verbal route descriptions provide a useful alternative for humans navigating in a natural environment. The semantics that emerge from such descriptions encompass several modelling abstractions that have been long studied by spatial cognition. However, a formal representation of navigation descriptions still remains a research challenge. The objective of the research presented in this paper is to provide a modelling approach for the description and fusion of several verbal route descriptions, and to identify the relevant places that emerge. A semantic spatial network is derived, thus generating a conceptual map that might be used for pedestrian navigation. The semantic spatial network is generated after application of a genetic algorithm and fusion rules to verbal route descriptions recorded by several humans navigating in a given natural environment. Preliminary results are encouraging but still have to be compared with real maps and with expert knowledge.