All-integer column generation for set partitioning: Basic principles and extensions

This paper presents a framework for such an all-integer column generation approach to set partitioning problems. We give the basic principles of all-integer pivots and all-integer column generation. We also state optimality conditions and introduce means for preserving a basis in the event that a heuristic is applied to the master problem. These extensions introduce flexibility in the design of a specific solution scheme of this kind, and with proper settings optimal or approximate solutions can be sought for.