Title of Presentation: Application of Middleware and Agent Technologies to a Representative Sensor Network

Primary (Corresponding) Author: John Kinnebrew

Organization of Primary Author: Vanderbilt University

Co-Authors: Nishanth Shankaran, Douglas Schmidt, Gautam Biswas, Dipa Suri

Abstract: Distributed real-time embedded (DRE) systems often operate in open environments where operating conditions, input workload, and resource availability cannot be accurately characterized a priori. Some DRE systems, such as the South East Alaska Monitoring Network for Science, Telecommunications, Education, and Research (SEAMONSTER) project, perform sequences of heterogeneous data collection, manipulation, and coordination tasks to meet specified objectives/goals. Open DRE systems are also required to operate with a high degree of local autonomy and adaptivity as new data is acquired and analyzed, as well as when environmental conditions change. Key challenges in managing open DRE systems include effective planning, scheduling, and online management of system resources to accommodate changing mission goals, environmental conditions, resource needs, and resource availability. This paper explores the benefits of an integrated planning, scheduling, and adaptive resource management framework that forms part of our larger multi-agent system MACRO. This integrated framework combines decision-theoretic planning with adaptive resource management to control and ensure efficient functioning of open DRE systems.