This paper develops a decentralized event-control strategy to avoid deadloc
k in automated manufacturing systems. On the basis of the characteristics o
f the part mix that is currently active, the system is partitioned in subsy
stems that can be controlled locally and independently using graph-theoreti
c tools. The controller of each subsystem can be chosen as the best comprom
ise between computational costs and flexibility in resource allocation and
depending on the particular layout of the subsystem. The decentralized sche
me allows the overall system to improve its flexibility and guarantee good
performance measures. The approach is particularly suitable for cellular ma
nufacturing systems that do not exclude intercell flow. In this case the me
thod determines the cells that can be controlled independently and the ones
that must be controlled jointly. A case study performed by discrete-event
simulation confirms the efficiency of the proposed methodology.