Routing problem with service choices

Abstract

This thesis finds solutions to the routing problem with service choices which is formulated as a capacitated minimum cost flow circulation problem with GUB constraints. The routing problem with service choices is solved using a specialized GUB branch and bound algorithm. Methods for node and GUB set selection are presented. A heuristic for finding good feasible solutions to initiate the branch and bound using vehicle size cuts is also derived. Furthermore, a network reduction scheme is formalized to reduce the size of the problem. This reduction is applied between pairs of nodes whose ground arcs have infinite upper-bounds. Initial experiments using the GUB branch and bound on several medium scale test problems appear promising. A variable tracking scheme which updates the status of the branching variables is included, which can be used to fully automate the branch and bound. This work supports the use of LP based GUB branch and bound for solving combinatorial problems with GUB constraints. Extensions to several related problems are also given.