Shop-floor control is known to be a hard problem because it must supervise
manufacturing operations execution according to plans, respect delays and d
eal with real-time system perturbations. Modern shop-floor control systems
must be reactive, i.e., be able to quickly and timely react to perturbation
s, be they external or internal perturbations to the system to be controlle
d.
This paper proposes a novel approach to shop-floor control. Especially, its
aim is to solve dynamic production control problems in real-time (due to w
orkstation breakdowns, urgent orders ...), to automate as much as possible
the control process in order to adapt the system to production plan modific
ations and to rationalize decision-making by a means of strong hierarchical
structure.
To achieve this, the system is based on a two-fold hybrid multi-agent platf
orm. First, control is hierarchically distributed and decision-making is ce
ntralized, thus this type of system mediates between centralized and hierar
chical architecture types. Centralization allows to avoid the competition b
etween agents for solving a problem while hierarchical distribution allows
each agent to take care of only one product. Second, a complete schedule is
completed at the beginning of the process, but it can be partially (or tot
ally) modified during the process. Thus, significant gains in terms of resp
onse times and reactivity capabilities are obtained thanks to this hybrid a
pproach.