Combining a logical and an analogical framework for route generation and description

Route generation is an ideal subject for studying the interactions of spatial and temporal reasoning in a practical setting. Starting from a cognitive study of pedestrian route descriptions in urban environments generated by human subjects, we identified the main requirements for a knowledgebased-system that would generate and describe routes in a cognitively plausible way. In this paper we present the main characteristics of the GRAAD system which provides an integrated logical and analogical framework for reasoning about space and time when generating and describing the routes of a virtual pedestrian (VP) in a virtual urban environment. The analogical framework is based on the notion of spatial conceptual map (SCM) which is used to specify a spatial environment composed of landmark objects and medium objects (streets, avenues, etc.) and to simulate the displacements of VP in that environment. A module uses this analogical framework to generate candidate paths between any two positions located on medium objects of the SCM. The logical framework is based on a possible worlds approach used to carry out temporal reasoning and to generate plans that are used to compute VP’s displacements in the SCM and to generate natural language descriptions of these displacements. These descriptions are cognitively plausible because the SCM’s medium objects are partitioned into areas which can be directly described using verbal expressions similar to those used by people when describing routes.