Controlling agent interactions in a tightly coupled multiagent framework
Myeong-Wuk Jang, Joongmin Choi, Sangkyu Park, Sooncheol Baeg, Gowang-Lo Lee, Young-Hwan Lim · 1996
Controlling interactions among agents in a multiagent system has been a challenging issue [4,6,10]. The types of interactions include requesting jobs, replying with answers, reporting control information such as activation status or error messages, and so on. The control of agent interactions is a difficult task mainly because agents are normally heterogeneous and lack the knowledge about other agents. In this loosely coupled system, approaches for controlling agent interactions have been focusing on developing mechanisms for determining who can solve which task. Efforts for standardizing inter-agent communication languages such as KQML belong to this category [3,4]. Another difficulty lies in the management of common ontology for vocabularies that are exchanged among agents. Namely, it is possible that the same word means different things (ambiguity), or different words mean the same thing in different agents (conflicts) among different agents [5]. In a tightly coupled multiagent fram...