Multi-Agent System Research Map

Multiagent Interactions

Agents are social beings in the sense that they have the possibility to initiate interactions with other agents. Research in the field of agent interactions covers the different aspects of communication at varying abstraction levels. Fields of interest are the agent communication foundations, conversations and cooperation.

Agent communication takes place on a more abstract level than this is the case with object-oriented communication. Agent communication is based on the speech act theory [Sea69]. Communication is treated as a way of acting, as certain kinds of natural language utterances have the characteristics of actions (called speech acts), because they imply a so called "rational effect". Speech acts are the foundation for agent communication languages (ACLs, eg. KQML [KQM92], FIPA-ACL [FIP02a]). The agent communication language defines the set of allowed speech acts and their associated semantics (e.g. [FIP02b]).

Besides the description of the communicative act the content of a message has to be represented in an adequate way. Several different knowledge representation languages have been developed (e.g. FIPA-SL [FIP02c], KIF [KSE95], CCL [WCF+00], RDF[W3C99]). To establish a common conceptual ground for the notions used in the communication several agents share an ontology which is a conceptualization of the world, and conatins the needed notions and their relationships.

A set of interrelated messages forms a conversation, in which agents play different roles to achieve their individual or shared goals. Different kinds of conversations exist for different purposes. For example, negotiations are seen as a promising approach for flexible dynamic task delegation and execution. Conversations in multi-agent systems can be based on interaction protocols that define all the possible plots of a conversation. Several protocols already have been standardized by the FIPA. They reach from very simple protocols, eg. the FIPA Request Protocol [FIP02d] to very complex ones, eg. the FIPA English Auction Protocol [FIP02e].

Agents are autonomous entities that try to accomplish their personal goals. Another facet occurs when considering agents that have to reach a common goal. Agents often have different and specialized abilities and need each other to fulfil the superordinated goal. Cooperative problem solving [Dur89] is concerned with the problems of how to decompose the goal to sub-problems, how to assign these sub-problems effectively to individuals, and how to assemble an aggregate solution [SmDa80]. Question of cooperation are closely related to aspects of the social structures in a multi-agent system.

[Dur88]	E. H. Durfee. Coordination of Distributed Problem Solvers. Kluwer Academic, Boston, MA. 1988.
[FIP02a]	FIPA ORG. FIPA ACL Message Structure Specification. Document no. SC00061G, 2002.
[FIP02b]	FIPA ORG. FIPA Communicative Act Library Specification. Document no. SC00037J, 2002.
[FIP02c]	FIPA ORG. FIPA SL Content Language Specification. Document no. SC00008I, 2002.
[FIP02d]	FIPA ORG. FIPA Request Interaction Protocol Specification. Document no. SC00026H, 2002.
[FIP02e]	FIPA ORG. FIPA English Auction Interaction Protocol Specification. Document no. XC00031F, 2002.
[KQM92]	KQML Advisory Group. An Overview of KQML: A Knowledge Query and Manipulation Language.
[KSE95]	Knowledge Sharing Effort. KIF Knowledge Interchange Format.
[Sea69]	J. R. Searle. Speech Acts: an Essay in the Philosophy of Language. Cambridge University Press. 1969.
[SmDa80]	R. G. Smith, R. Davis. Frameworks for cooperation and distributed problem solving. IEEE Transactions on Systems, Man and Cybernetics, 11(1). 1980.
[W3C99]	W3C. Resource Description Framework (RDF) Model and Syntax Specification.
[WCF+00]	S. Willmott, M. Calisti, B. Faltings, S. Macho-Gonzalez, O. Belakhdar, M. Torrens. CCL: Expressions of Choice in Agent Communication. The Fourth International Conference on MultiAgent Systems (ICMAS-2000), Held July 7-12, 2000, Boston MA, USA. pages 325-332.