Abstract

A model-based approach to the recognition of assembly process states
is presented.  Force information acquired in the
process is interpreted using state classifiers in order to estimate
the current state of the assembly process.  The classifiers of the
assembly process state are formulated by using the theory of
polyhedral convex cones, which allows us to treat unidirectional
geometric constraints in a systematic manner.  We develop a systematic
method for generating the classifiers automatically based on
geometric models of assembly parts.  Using the classifiers generated on
a computer, a recognition system can determine the process state from
sensory information.

First, a symbolic representation of assembly processes with respect to
contact states is addressed.  Kinematic and static behavior of
assembly parts at each contact state are then analyzed by applying the
theory of polyhedral convex cones.  State classifiers that
discriminate contact states are formulated by using the polyhedral
convex cones, which directly provide a set of discriminant functions.
To reduce real-time computations, the classifiers are simplified to a
minimum set by using reduction rules of polyhedral convex cones.  The
algorithm to generate the state classifiers is then implemented on a
computer to demonstrate the usefulness of this approach.