A multi-agent architecture for intelligent gathering systems

  • Authors:

  • David Camacho;Ricardo Aler;Daniel Borrajo;José M. Molina

  • Affiliations:

  • Departamento de Informática, Universidad Carlos III de Madrid, Leganés, Madrid, Spain;Departamento de Informática, Universidad Carlos III de Madrid, Leganés, Madrid, Spain;Departamento de Informática, Universidad Carlos III de Madrid, Leganés, Madrid, Spain;Departamento de Informática, Universidad Carlos III de Madrid, Leganés, Madrid, Spain

  • Venue:
  • AI Communications
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a model to define heterogeneous agents that solve problems by sharing the knowledge retrieved from the WEB, and cooperating among them. The control structure of those agents is based on a general purpose Multi-Agent architecture (SKELETONAGENT) based on a deliberative approach. Any agent in the architecture is built by means of several interrelated modules: control module, language and communication module, skills modules, knowledge base, yellow pages, etc. The control module uses an agenda to activate and coordinate the agent skills. This agenda handles actions from both the internal goals of the agent and from other agents in the environment. In the paper, we show a high level agent model, which is later instantiated to build a set of heterogeneous specialized agents. The paper describes how SKELETONAGENT has been used to implement different kinds of agents and a specialized Multi-Agent System (MAS). The implemented MAS, MAPWEB-ETOURISM, is the specific implementation of a general WEB gathering architecture, named MAPWEB, which extends SKELETONAGENT. MAPWEB has been designed to solve problems in WEB domain through the integration of information gathering and planning techniques. The MAPWEB-ETOURISM system has been applied to a specific WEB domain (e-tourism) which uses information gathered directly from several WEB sources (plane, train, and hotel companies) to solve travel problems. This paper shows how the proposed architecture allows to integrate the different agents tasks with AI techniques like planning to build a MAS which is able to gather and integrate information retrieved from the WEB to solve problems.