Agent-based combinatorial auctions in supply chains
- Author
- Abbaas, Omar
- Published
- [University Park, Pennsylvania] : Pennsylvania State University, 2022.
- Physical Description
- 1 electronic document
- Additional Creators
- Landry, Steven
Access Online
- etda.libraries.psu.edu , Connect to this object online.
- Graduate Program
- Restrictions on Access
- Restricted (PSU Only).
- Summary
- This research explores the decentralization of supply chain management using agent-based techniques to solve traditionally difficult optimization problems. These problems are decomposed into smaller and easier-to-solve subproblems on the agents' level. An agent here can represent a resource, supplier, buyer, etc. Auction mechanisms are used for communication and coordination between the different agents. There are three objectives for this research: (1) study and expand the literature on agent-based models with auction mechanisms in decentralized supply chains, (2) explore application areas for this approach in modern supply chains, and (3) propose models and algorithms for a selected sample of application areas, then implement the proposed methodologies and evaluate their performance. In Chapter 3, we provide an application of this approach in the joint order acceptance and job shop scheduling problem. A set of available jobs is provided. Each job has a profit, ready time, due date, and deadline, and consists of a set of operations that have precedence relationships. An operation has a processing time and may require multiple units of capacity per time unit to be completed. Jobs that deviate from their due dates incur earliness and tardiness penalties, resources could have more than one unit of capacity, and the facility layout can be of any type. The manufacturer has the option to reject any of the offered jobs in order to maximize the overall profit while satisfying the capacity constraints. We present a mathematical model for the problem, then use Lagrangian relaxation and combinatorial auctions to solve it. First, the problem is decomposed into a set of job-level scheduling subproblems. Each job is optimized individually without considering capacity constraints. Profitable jobs at the individual level submit their bids to an auctioneer in order to acquire combinations of resource capacity-time units. Then, the auctioneer records the profit upper bound, resolves capacity conflicts to reach a feasible solution using our provided feasibility restoration algorithm, records the profit lower bound, and updates the Lagrangian multipliers. Experimental results show that the proposed methodology is capable of solving large-sized problems with competitive outcomes. In Chapter 4, we address the multi-item multi-sourcing supplier selection and order allocation problem. Suppliers use an efficient procurement combinatorial auction format to submit bids to a single manufacturer. Each bid carries information about the supplier's cost structure and potential discounts. Instead of using the traditional and computationally expensive static bidding language or the existing restrictive flexible bidding language, we propose a new flexible procurement combinatorial auction bidding language and then prove that it allows for more efficient auction outcomes. A Mixed Integer Non-Linear Programming (MINLP) model is developed for the problem under the proposed bidding language considering purchasing, transportation, ordering, administrative, and holding costs. The manufacturer makes a set of finished products that experience price-sensitive demand rates and sells them to end consumers. We use the logit function to represent the finished products' price-sensitive demand rates. These considerations make the results of this study realistic with the potential to be used in practical applications. Also, we derive a set of necessary optimality conditions that must exist in at least one optimal solution. At the end of the chapter, several numerical examples are solved to illustrate the bidding language and the derived theorems. Chapter 5 proposes an iterative auction mechanism for the multi-item multi-sourcing supplier selection and order allocation problem. Suppliers use the efficient flexible bidding language proposed in Chapter 4 to submit bids to a single manufacturer. In these bids, suppliers offer two types of discounts: synergy discount for bundles of related items, and all-unit quantity discount. We design an iterative procurement combinatorial auction mechanism that aims to maximize the manufacturer's profit by revealing the suppliers' minimum acceptable selling prices. The mechanism starts with initial bids submitted by the suppliers. The manufacturer stores all the submitted bids throughout the auction in a database and makes them accessible to the participating suppliers. The manufacturer solves an MINLP model to determine the winning suppliers and allocate orders for the current iteration. Then the manufacturer shares the solution with the suppliers and gives them the opportunity to adjust their bids for future iterations. However, to guarantee progress toward the suppliers' minimum acceptable prices, the manufacturer asks any supplier planning to submit an updated bid to adjust the selling prices and allow the manufacturer's profit to increase by a certain profit improvement factor. We develop a new MINLP model that helps the suppliers to maximize their profits while satisfying the manufacturer's profit improvement constraint. If a supplier cannot find a feasible solution that will improve the manufacturer's profit while earning the desired minimum profit markup, this supplier does not submit new updated bids. If there are no new updated bids submitted, the manufacturer reduces the profit improvement factor. The mechanism continues until no new updated bids are submitted and the profit improvement factor reaches a pre-determined minimum threshold. Finally, numerical examples are solved to illustrate the auction mechanism.
- Other Subject(s)
- Genre(s)
- Dissertation Note
- Ph.D. Pennsylvania State University 2022.
- Technical Details
- The full text of the dissertation is available as an Adobe Acrobat .pdf file ; Adobe Acrobat Reader required to view the file.
View MARC record | catkey: 38970023