Market-Based Systems for Workflow Scheduling This proposal presents the first year of a plan of research to develop an agent and market-based approach for the control of complex production processes, especially those arising in white collar and service delivery systems. Such systems are characterized by their complexity and by the ability of the service providers to create alternative paths through the process by their actions. The goal of this research is create theory based on agent and market-based control schemes which will guide the implementation of intelligent workflow systems. With these market-based schemes, control is decentralized to various human and computer agents (e.g., databases, schedulers, models) that bid for the opportunity to work on various service delivery tasks. Specifically, we wish to develop a general theory for market-based control for complex production systems. Building upon the literature in service process design and scheduling in operations management, agent-based systems in computer science, mechanism design in economics, and optimization theory, this research will explicitly deal with the following issues that naturally arise in white-collar production systems, but are currently not dealt with in the literature: (a) the sequence-dependent nature of production systems that limits the use of standard auction/ bidding mechanisms (e.g., Vickrey and Ledyard-Groves); (b) the stochastic nature of many of these systems and the need for robust control schemes; (c) asynchronous communications between the agents in the system, and (d) the need for agents to learn and adapt to changing process conditions. In addition, a careful analytical analysis will be undertaken to ascertain the relative merits of such a market-based schema as compared to more traditional, centralized control systems such as optimization-based scheduling models. The ultimate goal of this research is to create a generic technology for adaptive intelligent workflow systems based on solid theoretical underpinnings and to test its advantages and limitations as compared with more traditional production/ resource scheduling methods. Through this research, a rigorous approach will be taken to model white-collar work as a collaborative effort between multiple people and computers. + _ Ûª? ÑOh ª' +'ª?0 à + ] $ H l + ¢ ? D h + R:WWUSERTEMPLATENORMAL.DOT
