Other qualitative spatial reasoning techniques have been proposed that vary widely from the combination of quantity space, system state, and envisionment shared by most of the systems described above.
Examples of these are Schmolze's ``Physics for Robots'' paper [Shm86], which tries to extend Hayes' ontology of liquids to the situations that would be encountered by a robot in a kitchen and Davis' ``Ontology of Physical Actions'' [Dav85], which has similar aims at a more general level - a formal theory of solid objects that humans encounter - for use by a robot. Davis also describes a representation that can be used for reasoning about spatial relationships between objects in terms of containment and relative location [Dav84a].
These examples are normally classified as ``Naive Physics'', since they attempt to find a formal basis for reasoning about very fundamental human concepts of space, rather than trying to solve any particular problem of motion or shape description. Work of this kind may become important to the type of qualitative reasoning described above if it is to be incorporated in a more general system for reasoning about action in the world. At this stage, work relating to qualitative spatial reasoning can proceed without requiring study of these issues.
Some use has also been made of qualitative methods for solution of real-world problems, because of the advantages of qualitative data as a basis for rule-based decision making. Two examples of this are a system for qualitative description of motion and position in an autonomous vehicle [BB87], and one for qualitative description of plug and socket shape when mechanical parts are to be fitted together [GB87]. These projects give examples of approaches to qualitative representation, but do not use qualitative methods (such as envisionment) for any type of planning or analysis. The representations developed for these systems tend to have less generality, since they are concerned only with distinguishing between specific situations or objects, rather than with general techniques for qualitative spatial reasoning.
The system which I describe in chapters 4 and 5 also departs in some measure from the mainstream of qualitative reasoning, but is more closely related to the mainstream than those of Schmolze, Davis, Burger and Bhanu, or Green. Its representation of state is more general than the ``mechanism world'' systems, in that a given state describes the complete relative positions of all objects in a scene (not just contacts). It uses a space description method that is related to the quantity space, but this is adapted so that Forbus's reasons for criticism of the quantity space in two dimensional situations are avoided. The main features of the approach, however, were initially developed from consideration of robotic reasoning and solid modelling systems, rather than from other qualitative reasoning projects.