Research on Lateral Transshipment in Supply Chain Inventory Systems
供應鏈庫存系統中橫向調貨研究
Student thesis: Doctoral Thesis
Author(s)
Detail(s)
Awarding Institution | |
---|---|
Supervisors/Advisors |
|
Award date | 19 Jun 2018 |
Link(s)
Permanent Link | https://scholars.cityu.edu.hk/en/theses/theses(493df799-f509-4558-8446-9fcd796f15fb).html |
---|---|
Other link(s) | Links |
Abstract
Effective inventory management is critical to improve supply chain performance and reduce operational cost. Imbalances exist extensively in supply chain systems due to the demand uncertainty in a dramatically changing market. In addition to the traditional vertical cooperation, more enterprises adopt horizontal cooperation to increase their profits. Lateral transshipment has attracted much attention from researchers and practitioners due to its potential in improving customer service while controlling total cost. Traditional research on lateral transshipment is limited by the availability of data and solution complexity, many models are developed based on simplified problems, which has limited their applications in reality. Based on the research gaps, and considering the easy access to real-time data, fast demand forecast update and customer's sensitivity to time in the information era, this dissertation devotes to solving the following three problems:
1. Propose a dynamic preventive transshipment policy based on the Markov decision process in a centralized inventory system. A heuristic algorithm combined with stochastic dynamic programming is developed to solve this problem. The dynamic method is able to take advantage of real-time inventory and demand data when determining the timing and quantity of transshipment, which has overcame the passive effect of static policy. Current research that applying dynamic method is limited to emergency transshipment, while the research on preventive transshipment is insufficient. The decisions for preventive transshipment are more difficult because of higher uncertainty. The problem scale becomes larger as the number of retailer and the size of demand become larger. This will lead to ``the curses of dimensionality" in solving the problem. Based on the characteristics of the system and the model, a three-step heuristic solution procedure is designed to solve the problem. First, a decomposition method is applied to decompose the multi-retailer problem into pairs of two-retailer problems. Second, a sorting heuristic is designed to narrow down the action space. Third, two dynamic programming approaches are designed to solve problems of different scales. Backward dynamic programming approach is used to obtain exact solutions for small-scaled problems. Approximate dynamic programming approach is used to obtain near optimal solutions for large-scaled problems. Extensive numerical instances prove the effectiveness of the dynamic policy and the heuristic algorithm. The results demonstrate that the proposed policy can improve the profitability of the system significantly. Moreover, the demand distribution of each retailer has an obvious impact on the effectiveness of transshipment.
2. Propose a replenishment and preventive transshipment policy in a decentralized two-retailer inventory system when updated demand information is considered, and prove the existence of unique Nash equilibrium order quantity. The updated demand forecast is more accurate and it has an impact on the transshipment and replenishment decisions. The replenishment decision is originally made with demand uncertainty. The forecast updates bring double uncertainties for the decision, which makes the solution procedure more complex. The demand forecast update is modeled mainly in the research on inventory systems with two/multiple replenishments. It has been ignored in the research considering transshipment. To the best of our knowledge, not any research has investigated the preventive transshipment problem with forecast updates. Therefore, this dissertation proposes a two-stage model to solve this problem. Backward induction is applied: First a threshold transshipment policy based on updated demand information is proposed at stage 2; then the existence of a unique Nash equilibrium order quantity for each retailer is proven at stage 1. A centralized system and a newsvendor-like system without transshipment are investigated as benchmarks to evaluate the performance of the proposed policy. Numerical examples are conducted for retailers with both identical and non-identical demands. The results demonstrate that the transshipment policy can greatly improve the system performance when the forecast updates are highly informative. This policy brings more benefit for the system with non-identical retailers. This conclusion can guide retailers in their search for transshipment partners in practical application.
3. Propose the optimal transshipment and replenishment policy in a decentralized two-retailer inventory system with partial backordering. Complete backordering or complete lost sale is modeled in most of the existing research when there is a shortage. Partial backordering as a more realistic phenomenon has seldom been considered in the research on transshipment. This dissertation makes further investigation on transshipment under partial backordering. Not only the independent emergency transshipment and independent preventive transshipment are analyzed, but also a combined policy is studied. First, a single-stage model is established for the system with emergency transshipment. The unique Nash equilibrium of order quantity is proven. Then, a two-stage model is developed for the system with preventive transshipment. The optimal preventive transshipment policy and the unique Nash equilibrium of order quantity are obtained. Next, a two-stage model is developed for the system with the combined policy. The transshipment quantity and order quantity are identified. The results from extensive numerical examples show that these three policies all can improve profit and flexibility of the system. The performance of a specific policy is determined by the feature of the system: the transshipment price, the location of the transfer point, the backorder cost and the fraction of backordering. None of these policies has absolute advantage in all systems. Partial backordering has an obvious effect on the quantity and profit of transshipment, which has brought new perspective for transshipment decisions. This research assists retailers in deciding which transshipment policy should be chosen.
Focusing on the transshipment and replenishment decisions in different supply chain inventory system, this dissertation enriches the theoretical research, and provides managerial implications for managers in different enterprises.
1. Propose a dynamic preventive transshipment policy based on the Markov decision process in a centralized inventory system. A heuristic algorithm combined with stochastic dynamic programming is developed to solve this problem. The dynamic method is able to take advantage of real-time inventory and demand data when determining the timing and quantity of transshipment, which has overcame the passive effect of static policy. Current research that applying dynamic method is limited to emergency transshipment, while the research on preventive transshipment is insufficient. The decisions for preventive transshipment are more difficult because of higher uncertainty. The problem scale becomes larger as the number of retailer and the size of demand become larger. This will lead to ``the curses of dimensionality" in solving the problem. Based on the characteristics of the system and the model, a three-step heuristic solution procedure is designed to solve the problem. First, a decomposition method is applied to decompose the multi-retailer problem into pairs of two-retailer problems. Second, a sorting heuristic is designed to narrow down the action space. Third, two dynamic programming approaches are designed to solve problems of different scales. Backward dynamic programming approach is used to obtain exact solutions for small-scaled problems. Approximate dynamic programming approach is used to obtain near optimal solutions for large-scaled problems. Extensive numerical instances prove the effectiveness of the dynamic policy and the heuristic algorithm. The results demonstrate that the proposed policy can improve the profitability of the system significantly. Moreover, the demand distribution of each retailer has an obvious impact on the effectiveness of transshipment.
2. Propose a replenishment and preventive transshipment policy in a decentralized two-retailer inventory system when updated demand information is considered, and prove the existence of unique Nash equilibrium order quantity. The updated demand forecast is more accurate and it has an impact on the transshipment and replenishment decisions. The replenishment decision is originally made with demand uncertainty. The forecast updates bring double uncertainties for the decision, which makes the solution procedure more complex. The demand forecast update is modeled mainly in the research on inventory systems with two/multiple replenishments. It has been ignored in the research considering transshipment. To the best of our knowledge, not any research has investigated the preventive transshipment problem with forecast updates. Therefore, this dissertation proposes a two-stage model to solve this problem. Backward induction is applied: First a threshold transshipment policy based on updated demand information is proposed at stage 2; then the existence of a unique Nash equilibrium order quantity for each retailer is proven at stage 1. A centralized system and a newsvendor-like system without transshipment are investigated as benchmarks to evaluate the performance of the proposed policy. Numerical examples are conducted for retailers with both identical and non-identical demands. The results demonstrate that the transshipment policy can greatly improve the system performance when the forecast updates are highly informative. This policy brings more benefit for the system with non-identical retailers. This conclusion can guide retailers in their search for transshipment partners in practical application.
3. Propose the optimal transshipment and replenishment policy in a decentralized two-retailer inventory system with partial backordering. Complete backordering or complete lost sale is modeled in most of the existing research when there is a shortage. Partial backordering as a more realistic phenomenon has seldom been considered in the research on transshipment. This dissertation makes further investigation on transshipment under partial backordering. Not only the independent emergency transshipment and independent preventive transshipment are analyzed, but also a combined policy is studied. First, a single-stage model is established for the system with emergency transshipment. The unique Nash equilibrium of order quantity is proven. Then, a two-stage model is developed for the system with preventive transshipment. The optimal preventive transshipment policy and the unique Nash equilibrium of order quantity are obtained. Next, a two-stage model is developed for the system with the combined policy. The transshipment quantity and order quantity are identified. The results from extensive numerical examples show that these three policies all can improve profit and flexibility of the system. The performance of a specific policy is determined by the feature of the system: the transshipment price, the location of the transfer point, the backorder cost and the fraction of backordering. None of these policies has absolute advantage in all systems. Partial backordering has an obvious effect on the quantity and profit of transshipment, which has brought new perspective for transshipment decisions. This research assists retailers in deciding which transshipment policy should be chosen.
Focusing on the transshipment and replenishment decisions in different supply chain inventory system, this dissertation enriches the theoretical research, and provides managerial implications for managers in different enterprises.
- Supply Chain, Inventory Management, Lateral Transshipment, Dynamic Programming, Nash Equilibrium