Article Preview
TopA Two-Stage Robust Optimization For Reliable Logistics Network Design Via Evolutionary Computation
Designing a logistics network is a critical decision-making problem in logistics management that involves determining the optimal number, location, and capacity of logistics facilities and allocating customer demand to these facilities (Cui et al., 2022; Zhang, 2022). The goal is to minimize logistics costs while satisfying customer service requirements and other operational constraints. An effective logistics network design is crucial for the success and competitiveness of logistics service providers because it directly affects the efficiency, reliability, and responsiveness of logistics operations. A poor network design can lead to excessive costs, delays, and customer dissatisfaction, whereas an optimal network design can improve service levels, reduce costs, and increase profitability. Therefore, logistics network design is a strategic priority for logistics managers who seek to optimize their operations and gain a competitive advantage in the market (Farahani et al., 2014).
Traditional logistics network design models often assume a deterministic and static environment that may not reflect the reality of logistics operations subject to various uncertainties and disruptions, such as natural disasters, equipment failures, and labor strikes. Ignoring these uncertainties and disruptions can lead to suboptimal or infeasible logistics network designs that are vulnerable to disruptions and fail to meet customers’ needs (Hou & Jia, 2023). In recent years, various logistics networks have been disrupted by natural disasters, pandemics, and man-made accidents, resulting in significant economic losses and customer dissatisfaction. For example, the COVID-19 pandemic has exposed the fragility of global supply chains and highlighted the need for more resilient and flexible logistics networks that can adapt to sudden changes in demand and supply. Similarly, the increasing frequency and severity of natural disasters, such as hurricanes, floods, and earthquakes, has demonstrated the importance of designing logistics networks that can withstand and recover from disruptions (Cui et al., 2016; Elluru et al., 2019; Ivanov, 2022). Therefore, it is essential to consider the uncertainties and disruptions of logistics network designs and develop reliable and resilient logistics networks that can adapt to changing environments and recover from disruptions (Govindan et al., 2018). This requires a paradigm shift from deterministic and static optimization models to stochastic and dynamic models that can capture the uncertainty and complexity of real-world logistics operations.
To address this challenge, robust optimization has emerged as a promising approach to designing logistics networks that are resilient with regard to uncertainties and disruptions (Adelhuette et al., 2023; Duy, 2023). Robust optimization aims to find feasible and near-optimal solutions under various uncertain parameters, thus providing protection against the worst-case scenarios (Zeng et al., 2023). However, traditional robust optimization models for a logistics network design often focus on a single decision-making stage, such as the strategic level of facility location or the tactical level of transportation planning, without considering the interactions and trade-offs between different stages. In practice, logistics network design involves multiple decision-making stages, such as strategic decisions on facility location and capacity, tactical decisions on transportation and inventory management, and operational decisions on-demand allocation and routing (Gulpınar et al., 2013). Ignoring the multistage nature of logistics network design may lead to suboptimal or infeasible solutions that do not account for the logistics network’s long-term impacts and short-term adjustments. Therefore, it is necessary to develop multistage robust optimization models that can integrate the different levels of decision making and capture the dynamics and uncertainties of logistics operations over time (Govindan et al., 2020; Sadjady & Davoudpour, 2012).