The Agent Metamodel in CATALINA (Cognitive AgenT prActicaL reasonINg Architecture)

Several recent studies explore the relevance of the different aspects of consciousness in agent reasoning, often starting from cognitive science studies. Besides, the agent community always found a relevant reference in the practical reasoning paradigm proposed by Bratman in his works. The proposed CATALINA architecture aims to blend contributions from these two worlds by implementing a refined attention mechanism together with the BDI reasoning. The core of CATALINA is the Global Workspace (GW), inspired by the Baars’ Global Workspace Theory. Agent’s brain functionalities are modularised as executive functions, taking inspiration from Buheler’s theory of the brain executive system. Among them, the Reasoner Function is implemented as a practical reasoner resembling Bratman’s architecture. Blending these theories poses several challenges, among them the definition of an underlying metamodel that ensures coherence and collaboration among the ideas coming from the inspiring theories. To evaluate the performance of CATALINA’s attentional and deliberative mechanisms, we developed a simulation of an autonomous vehicle navigating between cities that receives several desires from the designer and, in executing them, shifts its attention among different tasks.