Experience with interior point optimization software for a fuel planning application

Abstract

The Karmarkar interior point algorithm has made it possible to solve large-scale decision problems that previously could not be solved in reasonable time, or were too large to be solved at all. In this paper, the authors present the results obtained from using the KORBX Advanced Mathematical Programming System (KMPS), which uses Karmarkar's interior point algorithm, to solve a number of linear optimization problems arising from a long-term fuel planning problem. Comparison with the results obtained using software based on the simplex method demonstrates the drastic improvements in solution time for the interior point method, especially with increase in problem sizes. This confirms earlier comparisons of interior point and simplex methods. The paper includes preliminary ideas and results on ways to combine the interior point and simplex methods in order to benefit from the superior speed performance of the former, and the warm-start and hot-start capabilities of the latter methods. >