Improving Solutions of a Class of Allocation Problems by Cyclic Shifts of Resources

Abstract

A class of multi-resource allocation problems with non-linear return functions is considered, for which some dedicated numerical solution methods exist. The approximate solutions produced by these methods will typically not possess some minimal properties known to be present in the corresponding optimal solutions. The present paper shows how this defect can be corrected by a cyclical shift of resources in the solution matrix. A favourable side-effect is that the shifting process normally leads to an increase in the value of the objective function.