Ontology-based Reflective World Model for Autonomous Agents

01.03.2008 - 31.01.2011
Research funding project
Today's distributed physical systems, e.g., for automated manufacturing or logistics have to deal with increasing complexity and a need to adapt to rapidly changing environments. Traditional hierarchical and centralized approaches are not adequate for control of such distributed systems and can fail due to insufficient capabilities to cope with high degree of complexity and practical requirements for robustness and flexibility. A useful approach for decentralized control is to employ autonomous software agents. Based on their autonomous local behavior and cooperation with each other, they can achieve desired global behavior. This approach can be used in distributed physical systems by assigning software agents to physical components. Such a component can be viewed as an embodiment of its assigned software agent. Viewed from the other side, the software agent serves as an artificial "intelligence" of the physical component. The assigned software agent is supposed to guide the physical component to show useful local behavior and to cooperate with other such components. Based on the local behavior, the overall control of the distributed physical system would achieve desired global behavior. As an extension to the usual approaches for software agents, we propose to integrate a reflective symbolic "world" model based on ontologies into each software agent. Given an ontology of the "world" surrounding the agent and its assigned physical component, the agent is provided with a real-time representation of its environment built from sensor data and communication with other agents. In addition, this model integrates all the information that represents the physical component associated with the software agent. This information includes a symbolic representation of the "self" of this component and its relations to surrounding environment ("world"). In this new sense, the model becomes reflective. This reflection provides a clean representation of the relevant knowledge and information. We conjecture that it will facilitate the self-localization of the embodiment and cooperation with other components. In addition, it may facilitate the reasoning and the acting upon the "self".

People

Project leader

Project personnel

Institute

Contract/collaboration

  • Ing. Punzenberger COPA-DATA GmbH
  • Rockwell Automation s.r.o.

Grant funds

  • FFG - Österr. Forschungsförderungs- gesellschaft mbH (National) Austrian Research Promotion Agency (FFG)

Research focus

  • Distributed and Parallel Systems: 40%
  • Cognitive and adaptive Automation and Robotics: 20%
  • Computational Intelligence: 40%

Keywords

GermanEnglish
Software Agentensoftware agents
Semantische Systemesemantic systems

External partner

  • Institut für Automatisierungs- und Regelungstechnik

Publications