Norman F. Carver III
Ph.D. Computer Science, University of Massachusetts
Current Research Interests:
Multi-agent systems, distributed problem solving, sensor interpretation, knowledge-intensive control of AI systems.
Dr. Carver's research is mainly in the area of distributed problem solving (DPS). This is a subfield of multi-agent systems (MAS) that studies how to solve large-scale problems using distributed systems of intelligent software agents. Key issues include: the effects of problem decomposition and system organization on system performance, methods for designing coordination strategies that limit agent communication while still providing high-quality solutions, and the design of systems whose performance degrades gracefully as agents and/or communication links fail. Much of Carver's research has focused on distributed sensor networks, an application for DPS/MAS whose importance is increasing rapidly as it becomes practical to build networks of hundreds or even thousands of microsensors. His work involves both empirical and theoretical approaches, and is increasingly focused on methods for dealing with the effects of scale in very large agent systems. One long-term project is using a MAS testbed to develop empirical data that links the performance of various DPS strategies to different classes of sensor network problems. The goal is to identify useful classes of sensor network problems and build a library of DPS strategies that are appropriate for each. Much of the theoretical work has been based on the use of Decentralized Markov Decision Processes (DEC-MDPs) for modeling MAS problems and producing minimum communication coordination strategies. This work has received funding from two grants by the National Science Foundation.
- E. Khorasani, N. Carver, and S. Rahimi, "Performance Evaluation of DPS Coordination Strategies Modeled in Pi-calculus," International Journal of Intelligent Information and Database Systems, 2009.
- N. Carver, "Efficient Approximate Inference in Distributed Bayesian Networks for MAS-based Sensor Interpretation," Proceedings of the 7th International Conference on Autonomous Agents and Multiagent Systems(AAMAS08), 2008.
- V. Lesser, K. Decker, T. Wagner, N. Carver, et. al., "Evolution of the GPGP/TAEMS Domain-Independent Coordination Framework," International Journal of Autonomous Agents and Multi-Agent Systems , 2004.