2024 article
Multi-period fourth-party logistics network design from the viability perspective: a collaborative hyper-heuristic embedded with double-layer Q-learning algorithm
Zhang, Y., Huang, M., Gao, Z., Jiang, S., Fang, S.-C., & Wang, X. (2024, April 12). INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH.
author keywords: 4PL network design; viability; collaborative hyper-heuristic; double-layer Q-learning; stress tests
UN Sustainable Development Goal Categories
12. Responsible Consumption and Production
(Web of Science)
Source: Web Of Science
Added: April 29, 2024
In the 'new normal' setting of a mega-crisis, the viability becomes the driving force for the fourth party logistics (4PL) network design. In this paper, the viability is characterised in terms of agility, resilience and survival sustainability as the response to changes in demand, disruption and survivability. The fortification and recovery strategies are considered in possible disruptions at transfer centres and third-party logistics providers. A novel mixed integer non-linear programming model is proposed to obtain the multi-period 4PL network solution with minimum total cost under viability constraints. Considering the NP-hard characteristic of problem and the non-convex of proposed model, the hyper-heuristic algorithm is designed. To take advantage of both global optimality seeking and local search ability, a collaborative hyper-heuristic embedded with double-layer Q-learning (CHHDLQL) algorithm is proposed. The effectiveness and efficiency of the proposed algorithm is demonstrated by the promising numerical results. By stress-testing the existing network, appropriate adjustments to fortification and recovery strategies can effectively cope with changes in demand and disruption. Furthermore, the impact of 4PL strategy, fortification and recovery strategies, and viability constraints are investigated. The demand satisfaction, network resilience and capacity can be improved by adjusting agility, resilience and survival sustainability to influence different component network costs.