Seokcheon Lee - Research Summary

1. Research Overview Since Dr. Lee joined the School of Industrial Engineering at Purdue University in 2007, he and his research team have made significant efforts to develop novel control approaches to the large-scale complex systems. These endeavors are in particular from the multidisciplinary perspective, recognizing that such an approach is capable of producing the simple yet effective techniques that are applicable to large-scale, real-time decision-making problems. The principles behind the approaches are in general from the human and biological societies, e.g. social welfare/auction from economics, swarm intelligence from biology, and centrality from the study on complex networks. Those principles found in human or biological society are similarly applied to the complex systems, giving rise to globally desirable performance through local decisions and interactions. 2. Component-based Software Networks [1]-[6] As a result of technological advances, a typical type of software systems has emerged. A large number of distributed software components are networked together through a task flow structure, and each component may have alternative algorithms among which it can choose to process tasks. However, increased complexity and vulnerability to adverse events of such systems give rise to the need for more sophisticated yet scalable control mechanisms. In this study a control mechanism is designed to meet the need. First, stress environments are implicitly modeled by quantifying resource availability of the system through sensors. Second, a mathematical programming model is built with the resource availability incorporated and with the stability in system behavior assured. Third, a multi-tier auction market is designed for solving the programming model to distribute computation and communication overheads. By periodically opening the auction market, the system can achieve desirable performance adaptive to changing stress environments while assuring stability and scalability properties.