Predicate Projection in a Bimodal Spatial Reasoning System

Abstract

Spatial reasoning is a fundamental aspect of intelligent behavior, which cognitive architectures must address in a problem-independent way. Bimodal systems, employing both qualitative and quantitative representations of spatial information, are efficient and psychologically plausible means for spatial reasoning. Any such system must employ a translation from the qualitative level to the quantitative, where new objects (images) are created through the process of predicate projection. This translation has received little scrutiny. We examine this issue in the context of a bimodal spatial reasoning system integrated with a cognitive architecture (Soar). As part of this system, we define an expressive language for predicate projection that supports general and flexible image creation. We demonstrate this system on multiple spatial reasoning problems in the ORTS real-time strategy game environment.