Publications

Detailed Information

Sharing Logistics Resources in the E-commerce Supply Chain under Uncertainty : 불확실성을 고려한 이커머스 공급망에서의 물류 자원 공유

DC Field Value Language
dc.contributor.advisor문일경-
dc.contributor.author이준혁-
dc.date.accessioned2023-11-20T04:18:28Z-
dc.date.available2023-11-20T04:18:28Z-
dc.date.issued2023-
dc.identifier.other000000177475-
dc.identifier.urihttps://hdl.handle.net/10371/196337-
dc.identifier.urihttps://dcollection.snu.ac.kr/common/orgView/000000177475ko_KR
dc.description학위논문(박사) -- 서울대학교대학원 : 공과대학 산업공학과, 2023. 8. 문일경.-
dc.description.abstractWith the growth of both communications technology and contact-free delivery demand, e-commerce has grown significantly during the past few years. However, fierce competition and the inherent uncertainty in the e-commerce marketplace have made retailers suffer from high operations costs. Because of such circumstances, the concept of the sharing economy has been confirmed as an innovative business model that can answer the need for more flexible logistics. Therefore, we aim to develop decision-making frameworks considering logistics resources sharing under uncertainty. In this thesis, we address three problems in the supply chain management field: (1) the perishable inventory problem, (2) the supply chain network design, and (3) the omnichannel retail operations. In addition, we consider the three different services or strategies to share logistic resources in the abovementioned problems.

First, we address the perishable inventory problem considering transshipment and online-offline channel system. We present a Markov decision process model by accommodating key attributes of the online-offline channel system. We develop the hybrid deep reinforcement learning algorithm based on the soft actor-critic algorithm to overcome the curse of dimensionality in the large-scale Markov decision process. In addition, we found that transshipment substantially reduces the outdating cost by allowing the offline channel to make good use of the old products that will be discarded in the online channel, which is new to the literature.

Second, we propose the supply chain network design problem considering the on-demand warehousing system. We propose the two-stage stochastic programming model that captures the inherent uncertainties to formulate the presented problem. We solve the proposed model utilizing Sample average approximation combined with the Benders decomposition algorithm. Of particular note, we develop a method to generate effective initial cuts for improving the convergence speed of the Benders decomposition algorithm.

Third, we address the omnichannel retail operations considering the third-party platform channel. We propose the stochastic optimization model considering both the retailer's supply chain and the third-party platform's supply chain for omnichannel retail operations. To tackle the intractability of the stochastic optimization model, we propose a decomposition approach based on the two-phase robust optimization approach. The experimental results suggest that a decomposition approach is scalable to large-scale problems while maintaining its high solution quality.
-
dc.description.abstract통신 기술이 발전하고 비대면 수요가 증가함에 따라 이커머스 시장은 최근 몇 년 동안 크게 성장하였으며, 이커머스 유통업체의 수 또한 증가하였다. 하지만 이커머스 시장의 경쟁과열과 수요의 불확실성으로 발생하는 높은 운영 비용으로 인하여 많은 유통업체가 어려움을 겪고 있다. 이러한 어려움을 극복하는 방안으로, 공유 경제는 유연한 물류 운영을 위한 비즈니스 모델로 주목받고 있다. 본 학위논문에서는 불확실성하에서 물류 자원 공유를 고려한 의사결정 모델을 개발하는 것을 목표로 한다. 또한, 공급망 관리 연구 분야와 관련된 다음의 세 가지 문제를 다룬다: (1) 소멸성 상품의 재고 관리, (2) 공급망 네트워크 설계, (3) 주문, 할당 및 배송 결정. 그리고 제시된 문제들에서 물류 자원을 공유하기 위한 세 가지 서비스 및 전략을 고려한다.

첫 번째로, 환적과 온라인-오프라인 채널 시스템을 동시에 고려한 소멸성 상품의 재고 관리 문제를 다룬다. 온라인-오프라인 채널 시스템의 특성을 고려한 마르코프 의사결정 모델을 제시한다. 또한, 마르코프 의사결정 모델에서 발생할 수 있는 차원의 저주를 극복하기 위해 소프트 액터 크리틱 알고리즘에 기반한 하이브리드 심층 강화학습 알고리즘을 개발한다. 그리고 환적을 통해 제품의 폐기 비용을 줄일 수 있는 효과를 확인한다. 두 번째로, 온디맨드 창고 시스템을 고려한 공급망 네트워크 설계 문제를 다룬다. 불확실성을 고려하고 제안된 문제를 모형화하기 위해 2단계 추계적 수리 모델을 제시한다. 표본 평균 근사법과 벤더스 분해법을 결합하여 제안된 문제를 해결한다. 특히, 효과적인 초기 절단면을 생성하여 벤더스 분해법의 수렴 속도를 증가시키는 방법을 개발한다. 세 번째로, 제3자 플랫폼의 판매 채널을 고려한 주문, 할당 및 배송 문제를 다룬다. 유통업체의 공급망과 제3자 플랫폼 기업의 공급망을 동시에 고려한 추계적 최적화 모형을 고려한다. 추계적 최적화 모형을 다루기 힘든 어려움을 해결하기 위해, 2단계 강건 최적화에 기반한 분해 기법을 제안한다. 실험적 결과를 통해 개발된 분해 기법이 대규모 문제들에서도 좋은 성능을 보이는 것을 확인한다.
-
dc.description.tableofcontentsChapter 1 Introduction 1
1.1 Motivations 1
1.2 Sharing logistics resources 3
1.3 Optimization approaches under uncertainty 5
1.4 Contributions 8
1.5 Outline of the thesis 10
Chapter 2 A hybrid deep reinforcement learning approach for a proactive transshipment of fresh food in the online-offline channel system 13
2.1 Introduction 13
2.2 Literature review 18
2.2.1 Overview of perishable inventory management and lateral transshipment 18
2.2.2 Lateral transshipment for perishable products 19
2.2.3 Reinforcement learning approach for inventory management 24
2.3 Problem description and mathematical model 26
2.3.1 Lateral transshipment for fresh foods in the online-offline channel system (OOCS) 26
2.3.2 Markov decision process for the proposed lateral transshipment problem 34
2.4 Solution methodology: hybrid deep reinforcement learning (DRL)approach 37
2.4.1 Soft actor-critic algorithm 37
2.4.2 SAC for discrete action space and prioritized experience replay 41
2.4.3 Two acceleration methods in the hybrid DRL approach: SQLT policy and reward shaping 45
2.5 Computational experiments 51
2.5.1 Performance analysis of the developed hybrid DRL approach for real-world data set 51
2.5.2 Advantages of transshipment on profit in the OOCS 59
2.5.3 Analysis for saving effect of outdating cost because of transshipment and heterogeneous shelf life 65
2.5.4 Managerial insights 70
2.6 Summary 72
Chapter 3 E-commerce supply chain network design using on-demand warehousing system under uncertainty 74
3.1 Introduction 74
3.2 Literature review 76
3.2.1 Dynamic facility location and supply chain network design under uncertainty 77
3.2.2 Supply chain problems in the ODWS and distinctive features of this study 80
3.3 Problem description and mathematical model 86
3.3.1 The supply chain with the ODWS 86
3.3.2 The two-stage stochastic programming model 92
3.3.3 Compact formulation 98
3.4 Solution methodology: Sample average approximation combined with the Benders decomposition algorithm 101
3.4.1 Sample average approximation 102
3.4.2 Benders decomposition algorithm 105
3.4.3 Acceleration method 108
3.5 Computational experiments 111
3.5.1 Description of the test instances 112
3.5.2 Performance analysis of the proposed algorithms 114
3.5.3 Performance analysis of the stochastic solution 117
3.5.4 Effects of the ODWS on the supply chain 119
3.5.5 Managerial insights 126
3.6 Summary 129
Chapter 4 A decomposition approach for robust omnichannel retail operations considering the third-party platform channel 131
4.1 Introduction 131
4.2 Literature review 134
4.2.1 Omnichannel retail operations 134
4.2.2 Third-party platform channel 136
4.2.3 Robust optimization 139
4.3 Problem description and mathematical model 141
4.3.1 Problem description 141
4.3.2 Stochastic optimization model 149
4.4 A two-phase approach (TPA) based on robust optimization 154
4.4.1 Phase 1 of TPA 154
4.4.2 Phase 2 of TPA 160
4.5 A decomposition approach (DECOM) 161
4.5.1 Phase 1 of DECOM 162
4.5.2 Phase 2 of DECOM 167
4.6 Computational experiments 175
4.6.1 Experiment 1: Performance analysis in small problems under symmetric and asymmetric demand distributions 177
4.6.2 Experiment 2: Computational efficiency of DECOM in large-scale problems 181
4.6.3 Experiment 3: Performance analysis by varying the production capacity 185
4.6.4 Managerial insights 188
4.7 Summary 189
Chapter 5 Conclusions 191
5.1 Summary and contributions 191
5.2 Future research 193
Appendices 196
Appendix Chapter A Supplementary materials for Chapter 2 197
A.1 Information about hyperparameters of the hybrid DRL approach 197
A.2 Pseudocode for SACDPE 198
A.3 The reasons for using the existing data 199
A.4 Improvement effects of transshipment varying the unit transshipment cost parameter 201
A.5 Improvement effects of transshipment varying the shelf life of online and offline channels 202
Appendix Chapter B Supplementary materials for Chapter 3 203
B.1 Parameter information 203
B.2 Comparison of performance for solving SAA problems and computational results about the two types of lead time 203
Appendix Chapter C Supplementary materials for Chapter 4 206
C.1 PLDR for Phase 2 of TPA 206
C.2 Linear deterministic model of PLDR 210
C.3 Experimental results on asymmetric demand distributions and different production capacities 218
C.4 Shapes of symmetric and asymmetric demand distributions 220
Bibliography 221
국문초록 243
감사의 글 245
-
dc.format.extentxiv, 247-
dc.language.isoeng-
dc.publisher서울대학교 대학원-
dc.subjectLogistics resources sharing-
dc.subjectE-commerce-
dc.subjectSupply chain management-
dc.subjectReinforcement learning-
dc.subjectStochastic programming-
dc.subjectRobust optimization-
dc.subject물류 자원 공유-
dc.subject이커머스-
dc.subject공급망관리-
dc.subject강화학습-
dc.subject추계적 계획법-
dc.subject강건최적화-
dc.subject.ddc670.42-
dc.titleSharing Logistics Resources in the E-commerce Supply Chain under Uncertainty-
dc.title.alternative불확실성을 고려한 이커머스 공급망에서의 물류 자원 공유-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.AlternativeAuthorJunhyeok Lee-
dc.contributor.department공과대학 산업공학과-
dc.description.degree박사-
dc.date.awarded2023-08-
dc.contributor.major공급망 관리-
dc.identifier.uciI804:11032-000000177475-
dc.identifier.holdings000000000050▲000000000058▲000000177475▲-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share