Planning for assembly requires reasoning about various tools used by h
umans, robots, or other automation to manipulate, attach, and test par
ts and subassemblies. This paper presents a general framework to repre
sent and reason about geometric accessibility issues for a wide variet
y of such assembly tools. Central to the framework is a nse volume enc
oding a minimum space that must be free in an assembly state to apply
a given tool, and placement constraints on where that volume must be p
laced relative to the parts on which the tool acts. Determining whethe
r a tool can be applied in a given assembly state is then reduced to a
n instance of the FINDPLACE problem (Lozano-Perez, 1983). In addition,
we present more efficient methods to integrate the framework into ass
embly planning. For tools that are applied either before or after thei
r target parts are mated, one method preprocesses a single tool applic
ation for all possible states of assembly of a product in polynomial t
ime, reducing all later state-tool queries to evaluations of a simple
expression. For tools applied after their target parts are mated, a co
mplementary method guarantees polynomial-time assembly planning. We pr
esent a wide variety of tools that can be described adequately using t
he approach, and survey tool catalogs to determine coverage of standar
d tools. Finally, we describe an implementation of the approach in an
assembly planning system and experiments with a library of over one hu
ndred manual and robotic tools and several complex assemblies. (C) 199
8 Elsevier Science B.V.