Script-based Architecture for Multi-agent Robotic Systems

Abstract

This paper proposes a generic framework for developing multi-agent systems. Various agent models or interaction mechanisms can be systematically implemented and evaluated. A script specifies a multi-agent system by defining each agent as a Finite State Automata (FSA). By using the state transition rule as the building block, an agent as well as a multi-agent system can be developed efficiently. A set of rules composes a protocol set which represents some agent’s behavior. Various protocol sets can be developed for different agent models. The framework is further applied to. multi-agent robotic systems in which reactivity and interaction are more important issues. Hence, each protocol set is decomposed into several module sets by functionality. Advantages of the decomposition are modulization, reactivity, parallel performance, and module set sharing. In addition, multiple protocol sets can be implemented for dynamic adaptation. A cooperation protocol for the Object-Sorting Task is utilized to illustrate the framework.