Solving Large Multilevel Lot-Sizing Problems with an Effective Heuristic Algorithm Based on Segmentation

We have developed an effective heuristic algorithm based on a soft optimization approach for solving the multilevel lot-sizing problems in a series assembly product structure [Kaku et al., 2006 and 2008]. The heuristic algorithm segments the solution space and points the search direction in which a better solution may exist, by using the structure information of the multilevel lot-sizing problem, so that better performance can be achieved. In this paper, we extend the heuristic algorithm to the general case of multilevel lot sizing problems with time-invariant cost structures and no restrictive assumption on the product structure. Comparing with the results obtained by the genetic algorithm developed in [Dellaert et al., 2000], the effectiveness of the heuristic algorithm is shown in simulation computation experiments.