An Improved ILP System for Driver Scheduling

Mathematical programming approaches to driver scheduling have been reported at many previous workshops and have become the dominant approach to the problem. However the problem frequently is too large for mathematical programming to be able to guarantee an optimal schedule. TRACS II, developed at the University of Leeds, is one such mathematical programming-based scheduling system. Several improvements and alternative solution methods have now been incorporated into the mathematical programming component of the TRACS II system, including a column generation technique which implicitly considers many more valid shifts than standard linear programming approaches. All improvements and alternative strategies have been implemented into the mathematical programming component of TRACS II to allow different solution methods to be used where necessary, and to solve larger problems in a single pass, as well as to produce better solutions. Comparative results on real-world problems are presented.