Hierarchical Model of Multi-Agent System for Spatio-Temproal Rich Domains

A model of a multi-agent system (MAS) for spatio-temporal rich domains is described. It is intended to be used in a dynamic computer vision system. Hence, the hierarchical organisation of MAS is proposed, where agents performing identical or similar tasks are grouped into clusters. Several clusters of agents comprise a level of the organisation. Such n levels form the MAS structure. An agent is defined as n-tuple consisting of knowledge and reasoning about spatial and temporal relations, capability of perception and acting and it also has communication ability. An agent is modelled by using a modification of a Petri net. The communication among agents, as well as among clusters (and levels) is based on blackboard paradigm. An example that illustrates the use of the proposed model for planning in temporal rich domains is given.