Scheduling cargo trains using genetic algorithms

The scheduling problem examined involves the transportation of iron ore by train. A schedule for the trains in the system should consider two important criteria: feasibility and minimisation of costs. Feasibility implies that none of the trains conflicts with any other en route. The cost criterion includes minimising the costs associated with stopping (for example, due to wear and tear on the brakes) and the costs of delaying any trains. The latter two imply that it is desirable for a train to reach its destination in the shortest time possible. The problem at hand is particularly well suited to a genetic algorithmic formulation as it is an NP-hard problem and, hence, it is impossible in practice to use a constructive algorithm to obtain a solution. Various GA specifications are presented for an example extracted from a real system and the results obtained are compared against each other