Pengembangan Pendekatan Berbasis Kendala dalam Menyelesaikan Persoalan Multi-Period Single Source

Abstract

The multi-period single-sourcing problem in this study is viewed as finding an assignment model for the problem of assigning each retailer to each specific facility such as warehouses and inventory items at the start of planning, with the goal of minimizing the cost of assignment, inventory acquisition and ordering over time with respect to satisfying the demand of each customer within the limited production capacity of the facility. The case studied in this problem is the case of the placement of inventory items distributed to customers online which is seen as a non-polynomial problem or NP hard problem that requires a solution algorithm, and the algorithm we offer is a direct search algorithm to solve the multi period single sourcing problem. The proposed direct search algorithm is a Branch and Price algorithm developed for the Generalized Assignment Problem (GAP) to a much more complete class of problems, called CAP (Convex Assignment Problems). In particular, we generalize the strategy of separating nonbasic variables from their constraints, combined with using active constraint methods to solve Generalized Assignment Problems (GAPs) into Convex Assignment Problems. We then identify important subclusters of the problem, which contain many variants of the multi period single sourcing problem, as well as variants of the GAP. The final result we found is an active constraint-based multi period single sourcing model that can minimize the damage of the optimal integer solution to solve the convex MPSSP problem.