City Research Online

Abstract

In order to perform some tasks, agents need to interact with each other. Thus, a Multi-Agent System (MAS) is a system composed by several agents, capable of mutual interaction. Communication is a kind of interaction that allows agents to work more effectively by sharing knowledge and exchanging information. Thus, communication allow agents to make queries, transmit information, perform declarations and to commit themselves to execute an action. For agents to communicate, a method of sequencing messages is needed (conversations). For conversations to be successful, pragmatic principles to guide the linguistic interchange should be make available. These principles should not violate crucial properties of agency such as autonomy, heterogeneity and proactiveness. Various classes of agent communication languages (ACLs) have been proposed to handle these issues, but standardization is still a holy grail. We claim that a rethinking of the general principles on the foundations of ACLs is needed. More specifically, a redistribution of the role played by the semantics (speech acts) and pragmatics (protocols and policies) of ACLs will dissolve some of the most important problems currently affecting agent communication. Agent communication has traditionally focused on the semantics of speech acts, and many important advances have been done on that respect. But for some exceptions discussed later on the pragmatic component has often been the poor relative, consisting usually on low-level contextual free protocols that merely established the order in which speech acts may be used. This thesis aims to show how a high-level ACL pragmatics is crucial to facilitate the use of the semantic component in a variety of scenarios and a necessary step towards standardization. In the pragmatic turn for agent communication that we are proposing, ACL pragmatics will take the form of conversation norms. These principles can be specifically formulated by means of conversation protocols and policies that govern agents’ message interchange taking into account contextual factors that affect agents’ decisions. Once the theoretical issues are established, we ground the pragmatic principles in a computational model and study its applicability using a declarative programming language.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology > Computer Science School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads