A pragmatic BDI architecture

Abstract

We present a unifying perspective of the individual control layers of the agent architecture InteRRaP. InteRRaP aims at modeling autonomous resource-bounded agents that interact with each other in dynamic multiagent environments. InteRRaP implements a pragmatic Belief-Desire-Intention (BDI) architecture, where the agent's mental state is distributed over a set of layers. Based on the processes of situation recognition and planning and scheduling, a uniform description for each control layer — the behavior-based layer, the local planning layer, and the cooperative planning layer — is provided. We demonstrate various options for the design of interacting agents within this framework in an interacting robots application. The performance of different agent types in a multiagent environment is experimentally evaluated.