Heterogeneous Scheduling and Rostering

We have designed and implemented a distributed model for planning and scheduling of transports in a railway company which is usable also for other production planning problems. Our model emphasizes the use of an agent for the coordination of distinct sub-problems occurring in railway scheduling such as for instance the allocation of track resources to transports, the allocation of vehicles to transports, and the allocation of staa to perform the transportation tasks. Using constraints as the key technology, we are interested in nding interfacing principles to combine several solvers to allow users, for instance reasoning agents, to have a uniform view of the combined constraint store of heterogeneous solvers. Abstraction of constraint stores is an important technique used to enable translations between heterogeneous solvers and to improve performance of the overall combined problem of producing a railway schedule. By using abstractions of relevant representations we achieve a hierarchical system which by focusing attention on diierent levels of abstraction can achieve an acceptable solution faster than previous techniques.