Magazine article AI Magazine

Qualitative Spatial Reasoning about Sketch Maps

Magazine article AI Magazine

Qualitative Spatial Reasoning about Sketch Maps

Article excerpt

Maps are a ubiquitous tool for human geospatial reasoning. Computer support for geospatial reasoning often takes the form of geographic information systems (GIS)--sophisticated systems that combine computational geometry with database techniques to provide powerful abilities to manipulate and visualize vast quantities of digital terrain data. GISs are the computer-aided design (CAD) software of geospatial tasks. However, it is well known that in engineering, CAD software is not terribly useful for the early stages of design (conceptual design) where basic design choices are made and principles of operation are laid out before detailed design decisions are made. There appears to be a similar stage of thinking in geospatial tasks, where sketch maps are used to reason through a problem. By sketch maps, we mean compact spatial representations that express the key spatial features of a situation for the task at hand, abstracting away the mass of details that would otherwise obscure the relevant aspects. Sketch maps today are typically drawn by hand on paper.

For computers to become useful partners in geospatial problem solving, they need to be able to work with sketch maps just as people do. Just as qualitative reasoning has proven valuable in software supporting conceptual design in engineering, we claim that qualitative spatial reasoning (Forbus, Nielsen, and Faltings 1991; Glasgow, Chandrasekaran, and Narayarian 1995) is essential for working with sketch maps. This article describes the progress we have made in a specific geospatial domain--battlespace reasoning--towards this goal. Warfare, while a regrettable aspect of human existence, remains one of the most complex and most important kinds of task that people do. Planning a battle requires coordinating a complex array of people and equipment to achieve sometimes subtle goals, in situations where there is great uncertainty and danger. Terrain plays a crucial role in military reasoning, because it affects movement, it can provide cover and concealment, and it affects the operation of sensors. Thus geospatial reasoning must play a major role in generating and reasoning about battle plans, called courses of action. Figure 1 is a screen shot of a COA drawn with nuSketch Battlespace.


The introduction of digital media into military operations has been slow for several reasons. One major problem is that commanders are adamant about not wanting to use mice and menus; they sketch, and they want to interact with software via sketching, just as they interact with their people. Dealing with sketch maps is a necessity for creating performance support tools for military operations. Although most of our experience has been in military tasks, the situation seems similar in other human geospatial reasoning tasks (Egenhofer 1997).

This article describes the techniques we have developed for qualitative spatial reasoning about sketch maps. We start by reviewing our approach to sketching and nuSketch Battle-space, our battlespace sketching software that has been used in several successful experiments. Next we provide an overview of the spatial representations of sketches and glyphs and the processing architecture that handles spatial computations. Then we describe the computation of spatial relationships, including qualitative topology and Voronoi diagrams. Path finding and position finding, two key tasks, are discussed next. We describe how these techniques are combined with analogical processing to provide a simple form of enemy intent hypothesis generation. Finally, we discuss plans for future work.

The nuSketch Approach to Sketch Understanding

Sketching is a form of multimodal interaction, where participants use a combination of interactive drawing and language to provide high bandwidth communication. Sketching is especially effective in tasks that involve space, such as geospatial reasoning. While today's software is far from being as fluent as sketching with a person, progress in multimodal interfaces has produced interfaces that are significantly more natural than standard mice or menu systems (cf. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.