Interaction of purposeful agents that use different ontologies

Abstract

At CIC we have developed a model that enables multi-threaded agents that do not share the same ontology, to interact and interchange information among them. The behavior of each agent is defined in a high-level language with the following features: (1) Each agent and each interaction can be described by several sequences of instructions that can be executed concurrently. Some threads belong to an agent, others are inherited from the scripts which they play or perform. (2) Of all the threads, the agent must select which ones to execute, perhaps choosing between contradictory or incompatible threads.(3) The model allows communications between agents having different data dictionaries (ontologies), thus requiring conversion or matching among the primitives they use (§4). (4) Some of the threads can be partially executed, thus giving rise to the idea of a "degree of satisfaction" (§6.2.1).(5) The world on which the agents thrive suffers unexpected events (§3), to which some agents must react, throwing them out of their current behavior(s). The model, language, executing environment and interpreter are described. Some simple examples are presented. The model will be validated using test cases based on real situations like electronic commerce, product delivery [including embedding agents in hardware], and automatic filling of databases (§6.2.2). © Springer-Verlag Berlin Heidelberg 2000.