Abstract:
A cooperative dispatching approach is proposed for scheduling a flexible manufacturing cell (FMC) that is modeled as a 'm'-machine flowshop. Many of the current flowshop scheduling heuristics and algorithms are either inflexible or based on assumptions that are overly restrictive for the highly automated FMCs. Priority dispatching rules, on the other hand, are more flexible but their reliance on local data can frequently result in mediocre schedules, particularly in the case of flowshops. Cooperative dispatching combines heuristic qualities and the flexibility of dispatching rules in a distributed scheduling procedure that employs more global and real-time data to support dispatching decisions at the machines. A dispatching selection at any machine is reached collectively after consultation, through agents operating over a local area network, with the other machines in the cell. The consultation is initiated every time a machine needs to make a loading decision, and it takes the form of a poll that seeks a consensus regarding which of the candidate jobs should be selected, taking into consideration the performance criterion. Neural networks are available to assist the machines in formulating their replies when polled. The cooperative dispatching approach was tested in computer simulations and compared to traditional dispatching rules, for cases of both static and dynamic job arrivals. It performed consistently better than leading dispatching rules for three different criteria and in three routing configurations. Cooperative dispatching was also observed to be less sensitive than other dispatching rules to the amount of part overtaking permitted in the intermediate buffers, an issue that is relevant to FMCs which may have particular in-process buffer selection constraints stemming from automation hardware restrictions.
