R-CAST

From Wikipedia, the free encyclopedia

This article describes a highly specialized aspect of its subject.
Please help improve this article by adding more general information.

R-CAST is a multi-agent architecture that supports decision-making teams by anticipating information relevant to their decisions based on a shared mental model about the context of decision-makings (DM). More specifically, R-CAST agents share a computational representation of the RPD decision model, which serves as the shared DM process between agents and human decision makers. In addition to anticipating information needed for decision makings, R-CAST agents also collaborate to seek and fuse information in a distributed environment such as Service-oriented architecture. R-CAST is developed at the Intelligent Agents Laboratory in the College of Information Sciences and Technology at Pennsylvania State University, lead by Dr. John Yen.

The R-CAST architecture is component-based and reconfigurable. By selecting components suitable for an application, R-CAST can be configured into a wide range of agents: from simple reflex agents to RPD-enabled agents. Key components of R-CAST include the RPD model interpreter, the knowledge base and reasoning engine, the information manager, the process manager, the communication manager, the task manager, and adapters for various domains. The RPD model interpreter matches the current situation with known experinces, which are organized into a hierarchy. Missing cues relevant to the current decision are identified. The information manager uses the information dependency in the knowledge base to infer missing information that is relevant to the higher-level cues. The communication manager then contact agents that provide the missing information.

R-CAST agents have been used to develop decision-making aids for human teams. They have also be used to study team cognition and human-agent collaboration issues.

The figure below shows an R-CAST decision process model that is composed of a reasoning engine, an RPD model, a task manager, and a process interpreter. Each component has its own parameters that can be adjusted according to interpretation needs. Each component also has its own knowledge representation. To build a model, one has to (a) determine what components are involved to compose the model, (b) analyze tasks and elicit knowledge for each component, and, (c) determine how the components should be configured.

[edit] Publications

Xiaocong Fan, Bingjun Sun, Shuang Sun, Michael McNeese, and John Yen, RPD-Enabled Agents Teaming with Humans for Multi-Context Decision Making, AAMAS 2006

X. Fan, S. Sun, M. McNeese, and J. Yen, Extending Recognition-Primed Decision Model For Human-Agent Collaboration, In Proceedings of the Fourth International Joint Conference on Autonomous Agents and Multi Agent Systems, pp. 945-952, The Netherlands, July 25-29, 2005.

X. Fan, S. Sun, B. Sun, G. Airy, M. McNeese, J. Yen, Collaborative RPD-Enabled Agents Assisting the Three-Block Challenge in Command and Control in Complex and Urban Terrain, In Proceedings of 2005 Conference on Behavior Representation in Modeling and Simulation (BRIMS), pp. 113-123, Universal City, CA, May 16-19, 2005.

X. Fan, S. Sun, and J. Yen, On Shared Situation Awareness for Supporting Human Decision-Making Teams, In Proceedings of 2005 AAAI Spring Symposium on AI Technologies for Homeland Security, pp. 17-24, Stanford, CA, Mar. 2005.