The initial goal of this research, as mentioned in the introduction to the thesis, was to find a technique for reasoning about unfastening. All unfastening problems can be described at a basic level as the removal of a part in an assembly into a position where its motion is unconstrained. The sliding configuration and ``free space'' goals used for the above qualitative reasoning systems were chosen as first steps in solving unfastening problems.
The main distinction between unfastening and the kind of planning problem discussed above is that fasteners normally involve a locking mechanism which is interlocked with the object fastened. Real world fasteners usually rely also on distortion of the fastener or fastened object. The stress resulting from this distortion, combined with friction, prevents the fastener or fastened object from moving.
Reasoning about the separation of interlocked objects requires a combination of the sliding and path planning methods discussed above, together with some enhancements - for example, a coarse boundary description of interlocked objects shows their edges as overlapping (as in the ``two combs'' example in the last chapter). Reasoning about such interlocking objects would require a different approach to that used for coarse path planning, since there is no question of the whole object fitting between via points.
The extension of these techniques to real world fasteners would also require substantial additions to the qualitative representation used. Most real world fasteners involve three dimensional features, in addition to non-rigid behaviour. Another limitation is that motion in unfastening often includes rotation in addition to translation.
These problems have been noted by others (see especially Sedas and Talukdar [ST87]), and other treatments of fastening, whether using a qualitative approach or not, also tend to concentrate on two dimensional cases, without rotation, and assuming rigid objects. It is clear that the number of new issues involved in real world fastening mean that fastening analysis is considerably more complex than the type of tasks addressed by my implementation. It is interesting to note, however, that this qualitative technique has performance that is comparable to other systems addressing the same problem, while offering a number of advantages over those systems. These advantages are discussed further in the next chapter.