在網路盛行之現今，傳統集中式控制決策的方式將不敷需求，而以分散式人工智慧之方式來解決企業內與之間的問題將愈形重要。因此本研究針對企業間之訂單競爭與交易，以及企業內訂單加工作業之工作中心排程兩個情境來進行探討。 而在分散式人工智慧領域中，多重代理者系統提供一具分散式特性之環境架構，而代理者之間的互動機制將是多重代理者系統運作成功的重要因素。 本研究欲以多重代理者系統來建構分散式環境架構，以此一架構建立企業之間以及工作中心之間的虛擬市場；並設計多重代理者訂單拍賣與交易機制為代理者之間的互動機制，以此方式來探討解決上述兩情境之問題。本研究最終目的為提供未來建構企業網路市場與分散式排程系統之研究的參考。 Metcalfe law has been proven in the Internet world. Almost everything in the business world will be changed due to the fast development of the information technology, especially in the Internet sector. More and more firms are joined together or compete against each other in pursuing the optimal profit. However, the number of the firms in the market make the problem in optimal resource allocation (or competition) becomes untraceable. Distributed decision making is one of the most promising approach in dealing with this issue. In this research, we look into two phases of the problem in distributed environment. One is the order allocation among firms; the other is the task schedule among the work centers (machines) in any particular firm. In this research, an auction-based multi-agent system is built in two folds. One is dealing with the dynamic order allocation among firms. The firms will compete the order (competition) and dynamically swap the held order (cooperation) according to the utility value. The other is to schedule the tasks of each order in any particular firm based on the capability of the work centers. It is shown in this research that the auction-based multi-agent system can perform well in the massive electronic marketplace and heterogeneous work centers environment.
