A dynamic programming-based heuristic is used to solve the crew balancing problem for freight rail. The model finds an optimal solution to the basic crew balancing problem, then uses heuristic adjustment for special circumstances surrounding particularities in crew management.

In practice, model results are presented to the end user for approval or modification before their enactment. With accurate information, we estimate the potential for savings of 4% of total crew deadhead and heldaway costs. When subjected to inaccurate forecasts typically experienced, however, the deterministic model performs worse than current practices.

We make suggestions for dealing with the uncertainties of forecasted train arrival and departure times in order to produce more implementable results.