One of the criticisms of grid-based intralogistics systems, and GridFlow in particular, has been: “What happens if a unit-conveyor fails?” Because individual cells control their behavior based on the states of their neighbors, would an individual failure significantly degrade performance of the entire system? Is there a way to modify the control scheme to minimize the effect of such failures?
In her recently completed diploma thesis, Zäzilia Seibold from the Institute for Conveying Technology and Logistics at KIT has modified the decentralized control rules in GridFlow to account for complete and partial failures of a unit-conveyor. Complete failure means the conveyor is unable to respond to messages from any direction and is unable to convey. A partial failure could mean an inability to respond to messages on one or more sides.
In the new scheme, conveyor modules detect that their requests are being ignored by the failed module, and they then send requests to, for example, move their items left or right around the failed module. In extreme cases, an otherwise not-requested item may be “pushed out” simply to clear up a traffic jam.
In the video below black items have been requested; gray items are stored; items with black circles are returning to the grid and seeking their appropriate home rows. A red X indicates a failed module; red lines indicate broken communication ports. A green circle indicates an item that has been artificially pushed out to work around a failed module.
The main conclusion of her thesis is that failures can be easily accommodated (well, it looks easy, anyway!), and that degradation in the system is very slight unless modules are highly prone to failure. In practice, the failed module could be changed out at the end of the day or shift. You can read details in Zäzilia’s thesis or in our conference paper.