Multi-agent systems at

SRI International's Artificial Intelligence Center

At SRI International's Artificial Intelligence Center, we have been pursuing the development of multi-agent systems within the context of the Open Agent Architecture (OAA), a framework for integrating a community of automated agents in a distributed environment.

Our research addresses a number of issues related to developing agent-based systems: defining an architecture enabling cooperation and communication among a dynamically evolving set of agents; supporting natural forms of interactions among humans and automated agents; allowing humans to collaborate with each other as well as with other automated agents; providing tools to facilitate the development, maintenance and distribution of new agents [3]; and applying these technologies to create agent-based applications.

The architecture we chose for agent interaction uses a federated approach, where facilitator agents provide scheduling and maintenance of communication among "client" agents during distributed communication [2].

The primary task of the facilitator is to decompose Interagent Communication Language (ICL) expressions and to distribute them to agents who have indicated a capability in resolving them. The OAA's ICL is similar to Stanford's ACL (KIF/KQML) agent language, possessing both a speech act (e.g. performative) layer and a message content layer. Our ICL, based on an extension of Prolog, has been designed with the objective of interoperating with several classes of natural language systems, in order to allow humans to easily issue commands to automated agents.

Interactions between human users and the agent community has been a priority in most OAA-related projects [4]. Most OAA applications permit multimodal combinations of speech, pen (handwriting & pen gestures) and GUI input, and produce responses to queries using graphical, textual and audio output, as a function of the resources of the user's machine. Many applications allow humans to synchronously collaborate with other human users, as well as with the automated agents comprising the system.

Approximately fifteen agent-based prototype applications have currently been implemented within the OAA framework. These include a set of "office assistant" agents which help a remote user access and manage resources such as calendars, databases and email from a telephone or PDA; a travel planning application where users can combine pen and voice input to make database requests of an interactive map in a natural manner [1]; and an educational application allowing students to collaborate with each other and with automated agents in the domain of the GLOBE* program. SRI will also be participating in AAAI's robot competition, using the OAA to provide coordination, communication and services for multiple robots as they work together on a common task.

References

1. A. Cheyer and L. Julia, "Multimodal maps: An agent-based approach," in Proc. of the International Conference on Cooperative Multimodal Communication (CMC/95), (Eindhoven, The Netherlands), May 1995.

1. P. R. Cohen, A. J. Cheyer, M. Wang, and S. C. Baeg, "An open agent architecture," in AAAI Spring Symposium, pp. 1-8, March 1994.

1. D. L. Martin, A. Cheyer, and G. L. Lee, "Agent development tools for the open agent architecture," in Proceedings of the First International Conference on the Practical Application of Intelligent Agents and Multi-Agent Technology, (London), pp. 387-404, The Practical Application Company Ltd., April 1996.

1. D. B. Moran and A. J. Cheyer, "Intelligent agent-based user interfaces," in Proceedings of International Workshop on Human Interface Technology 95 (IWHIT'95), (Aizu-Wakamatsu, Fukushima, Japan), pp. 7--10, The University of Aizu, 12-13 October 1995.

* Global Learning and Observations to Benefit the Environment (GLOBE).

http://www.globe.gov/

Back to FIPA Opening Forum program