Flexible job-shop scheduling becomes harder when a limited number of vehicles move jobs between stations. In that setting, the schedule itself determines how long jobs wait for transport, making the problem resource-coupled. A new arXiv paper introduces BOLT, a bound-oriented learning method designed for this transport-coupled scheduling problem.
BOLT builds on the idea of using bounds to steer learning, though the abstract leaves the exact architecture unspecified. The key contribution is recognizing that transport waits are not external constraints but outcomes of scheduling choices. By incorporating this coupling, BOLT aims to produce schedules that account for both machine timing and vehicle availability.
The work is presented as a single approach, so there are no competing claims to compare. The abstract positions BOLT as a way to handle the added complexity of scarce transport resources in flexible job-shop environments.