The Curse of Unintended Rounding Error: A Case from the Vehicle Scheduling Literature

In their recent paper, which was published in the Journal of the Operational Research Society, Cheshire et al. describe "A dual heuristic for vehicle scheduling". The authors claim, on the basis of 17 academic test problems from the vehicle scheduling literature, that "in most cases the new results are comparable with those of Foster and Ryan, whose technique showed improvements over all previous methods". But a careful examination of the relevant source materials shows that one subset of authors, which includes the authors mentioned here, forced possibly premature rounding upon intermediate results in their calculations. Thus the comparisons performed by Cheshire et al., and their conclusions, are suspect in principle. They are later shown to be rather misleading in practice.