Dynamic Scheduling Mechanism Design in Multi-Agent Systems

本研究计画主要探讨多代理系统之动态排程机制设计。多智能体系统调度可以被看作是基于互联网的供应链（e-Supply Chain）管理，制造，信息处理和其他服务行业的背景下出现的分散调度问题的抽象。排序问题是一个具有重要实际意义的研究领域。它是一项核心技术，能够在当今竞争激烈的全球市场中实现基于时间的竞争。 我们的长期目标是推进MAS调度的理论基础，并为开发MAS调度系统在供应链管理，制造和服务应用提供工程工具。在接下来的五年中，我们的主要研究工作包括：（1）设计一种用于动态MAS调度的回溯迭代拍卖机制：这种基于拍卖的调度机制将在博弈论环境下实现反应式调度;（2）设计动态投标语言：这些语言将增强先前制定的要求的表达能力-（3）响应性获胜者确定算法的设计：设计最优和近似赢家确定算法，以保证动态MAS调度的响应性。 该研究计划的成果将为那些固有的分散工业应用设计自动化调度系统提供理论和工程基础。我们预计这项研究将有助于大规模基于互联网的交易系统中动态调度组件的开发，并增强加拿大工业基于时间的竞争优势。从学术研究的角度来看，试图解决调度问题在一个动态的战略设置，我们解决算法和战略的挑战，在接口的机制设计，调度理论。这项研究计划将丰富现有的复杂性问题的理解，在应用经济模型的调度域。它还推进了调度研究，将经典的集中式模型扩展到分散和动态模型。

This research program is concerned with the design of dynamic scheduling mechanisms for multi-agent systems. Multi-agent systems scheduling can be seen as an abstraction of the decentralized scheduling problems arising in the context of Internet-based supply chain (e-Supply Chain) management, manufacturing, information processing and other service industries. Scheduling is an important research field of great practical interests. It is a core technology that enables time-based competition in today’s highly competitive global markets. Our long term objective is to advance the theoretical foundation of MAS scheduling and provide engineering tools for developing MAS scheduling systems in supply chain management, manufacturing and services applications. In the period of the next five years, our main research activities include: (1) design of a backtracking iterative auction for dynamic MAS scheduling: This auction-based scheduling mechanism will enable reactive scheduling in game theoretic setting; (2) design of dynamic bidding Languages: The languages will enhance the expressiveness of previously developed requirement-based bidding languages for MAS scheduling to represent agents’ backtracking preferences; and (3) design of responsive winner determination: We will design optimal and approximate winner determination algorithms to ensure responsiveness in dynamic MAS scheduling. The outcomes of this research program will provide theoretical and engineering foundations for designing automated scheduling systems for those inherently decentralized industrial application. We anticipate this research to contribute to the development of dynamic scheduling components in large scale internet-based trading systems and strengthen the time-based competitive advantage of Canadian industries. From academic research perspective, attempting to tackle scheduling problems in a dynamic strategic setting, we address algorithmic and strategic challenges at the interface of mechanism design, and scheduling theory. This research program will enrich the existing understanding of complexity issues in applying economic models to the domain of scheduling. It also advances scheduling research in terms of extending classic centralized models to decentralized and dynamic ones.