Specifying knowledge in cognitive multiagent systems using a class of hierarchical petri nets

Abstract

Nowadays, problems demands new computational paradigms to solve them. New computational systems must consider features which before was not. Among these new features is the distributed nature of some systems. Nevertheless, even problems with a centralized nature has so high complexity that the best solution is to break them in small blocks. To deal with this problems, multiagent systems (SMA) and agents are seen as an excellent alternative as a framework to model, specify and codify solutions. However, there is not a standard framework to aid in building such systems. In this paper, it is proposed a framework based on a specific Petri Net to model, specify and codify the knowledge since the social level of the system until the agents reactive levels. This approach is based on a explicit separation between the knowledge level and the mechanisms which will manipulate it. The Petri net model proposed presents an implied specification language which permit to deal with any knowledge representation formalisms, since production rules, ontology or even database structures. Besides that, the Petri net proposed has tools to analyse and validate the system about issues like redundancy, deadlocks and conditions associated to agents tasks.