16 December 2016
Linfox trials the MiniZinc platform to help make the right choices for multi resource transport problems.
When a ship docks with 20,000 containers, there are a lot of choices to be made.
Each container must end up on a train or truck, but the port can't fit all the trains and trucks at the same time. The ship must also be loaded for its next trip, and the berth must be cleared for the next ship. The containers therefore need to be unloaded as quickly as possible and piled up on the quayside. If the truck arrives for the bottom container first, things get awkward!
Operations Research was developed to marshal resources during wartime. Since then, many large software systems have been built to support the management of complex organisations such as airlines and port operations.
Over the years some of these systems have been released as commercial products, fronted with attractive user interfaces and integrated with popular business software. Like all software systems, they grow larger, more complicated and less flexible with each passing year.
As a result, the decision support software used by many companies has not kept up with their business needs, and they are incurring high software costs for solutions that too often fail to make the right choices .
The Monash MiniZinc platform, developed in collaboration with Data61/CSIRO, makes it easy to model and solve optimisation problems.
The user states the problem precisely and then chooses which optimisation technique to use from a menu. The platform does the rest. Development time has been reduced by an order of magnitude, and advances in optimisation enable MiniZinc to support solutions that are both flexible and scalable.
MiniZinc is used to schedule ships and tugs at Port Hedland, meeting various constraints such as the ship draft, ship speed and headway, different berth locations and tug availability.
The MiniZinc technology is the basis for an optimisation platform that Monash and Linfox are collaborating on to solve complex multi-resource transport problems.
This could be applied to common routing problems particular to the retail network, as the systems organise data to produce the least effort solution (whether measured in kilometres, hours, or cost) required to move goods .