A System for Qualitative Spatio-Temporal Reasoning

* supported by NSF Award CNS-0959958 Abstract The ability to reason about the spatial relationships between objects in the environment is a core requirement of mobile robots and many other types of intelligent systems. However, since objects frequently change their location, the representation of spatial location critically depends on time. Thus, what is really required is spatiotemporal reasoning. This paper presents a unified system for qualitative spatio-temporal reasoning consisting of representations for spatio-temporal knowledge and a spatiotemporal inference system to manipulate those representations. Following Almeida [2], we treat spatial relations as event types incorporating the interval-based temporal ontology of Allen [1] and the region-based spatial ontology of Randell, Cui & Cohn [13]. Our basic reasoning system is essentially an extension of the temporal reasoning system of Allen [1]. Critical to the success of this system is the ability to create and reason with maximal temporal intersections.