Coordinating Multiple Decision Makers in a Service Environment

该项目的目标是开发控制随机动态系统的方法，该系统由多个具有私人信息，不同性能目标和不同风险态度的交互代理填充。这项研究将开发基于收入分享/转让价格合同的机制，作为修改代理激励和诱导分散决策的一种方式，这些决策也是集中优化的。计算最优和次优收入共享合同的方法将使用近似动态规划的思想。本课程将研究多代理系统在运筹学和管理科学中的应用，包括航空公司联盟的协调、多代理风险管理和多代理服务系统，如果成功的话，本研究的结果将使我们更好地理解在随机动态系统中用于协调相互作用的分散代理的机制。它还将导致开发用于计算用于协调这些决策者的最优和/或次优收入共享合同的有效方法。这些结果将为组织和管理由多个代理填充的复杂系统提供指导。

The goal of this project is to develop methods for controlling stochastic dynamic systems that are populated with multiple interacting agents endowed with private information, different performance objectives and different risk attitudes. The research will develop mechanisms based on revenue sharing/transfer price contracts as a way to modify agent incentives and to induce decentralized decisions that are also centrally optimal. Methods for computing both optimal as well as suboptimal revenue sharing contracts will be developed using ideas from approximate dynamic programming. Applications to multi-agent systems of interest to operations research and management science including the coordination of airline alliances, multi-agent risk management, and multi-agent service systems will be studied.If successful, the results of this research will lead to a greater understanding of the mechanisms that can be used to coordinate interacting decentralized agents in stochastic dynamic systems. It will also lead to the development of efficient methods for computing optimal and/or suboptimal revenue sharing contracts that are used to coordinate these decision makers. These results will provide guidelines for organizing and managing complex systems that are populated with multiple agents.