Understanding dynamic agent's reasoning

  • Authors:

  • Nuno Lau;Luís Paulo Reis;João Certo

  • Affiliations:

  • Informatics Electronics and Telecommunications Department, University of Aveiro, Portugal;Artificial Intelligence and Computer Science Lab., University of Porto, Portugal;Informatics Electronics and Telecommunications Department, University of Aveiro, Portugal and Artificial Intelligence and Computer Science Lab., University of Porto, Portugal

  • Venue:
  • EPIA'07 Proceedings of the aritficial intelligence 13th Portuguese conference on Progress in artificial intelligence
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Heterogeneous agents that execute in dynamic, uncertain, partially cooperative, partially adversely environments have to take their decisions rapidly with an incomplete knowledge of actual environment conditions. This paper discusses different level of abstractions in agent's development for this type of domain, explains the principles of offline debugging and employs these principles in robotic agent's teams through the use of a new visual debugging tool. We argue that during the development of such complex agents, understanding agent reasoning is of crucial importance to the developer and that such understanding can only be done, in most of the cases, using an offline analysis of the agent's decisions. In order for the developer to rapidly perceive agent's reasoning we advocate visual debugging of the agent knowledge and reasoning at several levels of abstraction and at several different functional views. These principles have been applied with success in the context of a robotic soccer 2D simulation league team through the development of tools and extensive use of these analysis principles.