From interoperability to cooperation: Building intelligent agents on middleware (original) (raw)

1998, Lecture Notes in Computer Science

As agent technologies are increasingly being involved in telecommunication-related applications, the need for open standards is becoming critical. During the past years, different scientific communities gave birth to different standardization actions, such as the Foundation for Physical Intelligent Agents (FIPA) and the Object Management Group's MASIF (Mobile Agent System Interoperability Facilities). Although they finally share some major targets, the OMG and FIPA current results show their distinct origins, respectively with a Distributed Artificial Intelligence and Multi-Agent Systems awareness on the one hand, and a telecommunication and information technologies background on the other hand. In a context where these two actions think about joining their achievements to upgrade each other, this article reports several experiments, carried out during the five past years in the agent platforms field, mixing both the intelligence and the middleware aspects. 1.2 Multi-Agent Platforms Many multi-agent platforms offer modelling and implementation solutions to the distribution of intelligence. To a certain extent, one may consider multi-expert systems and Distributed Artificial Intelligence as parts of the origins of the multi-agent systems. Following this point of view, we can see through the introduction, in typical centralized Artificial Intelligence languages, of low-level communication features (e.g. TCP/IP sockets in Lisp or Prolog), or more elaborate communication structures (e.g. blackboards and Linda Interactor in [20]), the emergence of the first multi-agent platforms. Then, sophisticated models (actors [1]) and techniques (constraints, reflexivity [10]) have been merged to enhance multi-agent platforms. But, as the enhancements are going on, the resulting diversity and heterogeneity makes it difficult for a standard to emerge, besides the AI classics, which industry is just beginning to appropriate to itself. Moreover, when it comes to distribution and communication, specific solutions are often applied, sometimes through simulation. But, the more these platforms integrate to standards, in real applications, the more striking is the proof of their accuracy. Then, it seems useful to find a way of defining a standards-based bridge between the real applications and multi-agent platforms, without preventing their evolutions, but enforcing reusability and interoperability.