An Approach to Specify Knowledge in Multi-agent Systems Using Petri Nets

The use of approaches based on artificial intelligence, specially agents and multi-agents systems, permits to clearly distinguish the aspects of implementation and the knowledge that gives substrate to the system. Concerning the implementation aspects, there are some traditional frameworks and languages as AUML, Gaia, and JASON. However, this technologies are not comfortable in acquiring and structuring knowledge. In this paper, it is presented a Petri net model developed to specify knowledge in agents and multi-agent systems independently of frameworks and knowledge representation formalisms. The Petri net model presented permits to map the knowledge acquired and structured in any formalism and framework used to implement a computational system. Besides that, the Petri net tool permits to analyze and validate the knowledge elicited concerning aspects as redundancy, deadlocks and conditions associated to agent tasks. The main contribution of this approach is to shift the project focus to the knowledge level of the system.