Clustering-based NSGA-II on Multi-Objective Location-Routing Problem

Abstract

Determining facility locations and planning routes for efficient goods/materials distribution in the supply chain, known as the location-routing problem (LRP), poses a significant challenge in logistics. This research focuses on developing a method to achieve desirable outcomes in LRP with soft time windows by integrating clustering techniques with the NSGA-II algorithm. A case study is conducted to optimize commodities distribution in D.I Yogyakarta, Indonesia using the proposed method. The study demonstrates how clustering can enhance the results of the multi-objective location-routing problem with time windows (MLRPTW) through analysis. The solution set reveals that the clustering-based NSGA-II outperforms the classical approach in terms of both objective functions (cost minimization and service level maximization). By grouping retailers based on similarities in location and opening time windows, the method improved population initialization, resulting lower distribution center (DC) costs and higher retailer satisfaction. Retailers within the same cluster tend to share the same DC, which is closer in proximity, also tend to have similarity in terms of working hours (time windows). Moreover, incorporating clustering within NSGA-II decreases the computational time which is preferable by decision maker. We also explored and compared k-medoids as a clustering method alternative, the result shows there is no significant improvement compared to k-means.