Message-Based Agent Communications in a Tightly Coupled Multiagent System

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 (conflicts) in different agents (5). This paper presents methods for controlling interactions among agents in a tightly coupled multiagent framework. In this framework, an agent is aware of what agents are available, which is able to solve a query, and how to contact them. This characteristic leads to the simplicity in controlling both local and remote interactions among agents. The form of the communication packet used in information exchange is predetermined and transparent to all agents. Agent interactions are described within the platform of MASCOT that is a tightly coupled multiagent system developed for the role of a computer secretary. Protocols for message exchange are presented, and a scenario is given to demonstrate how the communication mechanism controls agent interactions in MASCOT. Compared to this, we present a mechanism for controlling interactions among agents in a tightly coupled multiagent framework . In this framework, an agent is aware of what agents are available, which is able to solve a query, and how to contact them, albeit the heterogeneity still remains. In our viewpoint, a tightly coupled multiagent system is built in a top-down manner by disassembling a monolithic agent into a set of task-specific agents, as opposed to a loosely coupled system that is the bottom-up integration of separate agents which already exist or are newly built (7). The main characteristics in the tightly coupled multiagent system are simplicity and efficiency in the control of both local and remote interactions among agents. The form of the communication packet used in information exchange is predetermined and transparent to all agents. The communication packet includes the control part that specifies the sender, the receiver, and the type of the message, and the data part specifies the action and the content. Although a simple point-to- point protocol might be adopted for this tightly coupled framework, it has a disadvantage that the communication module of each agent ought to take care of synchronization of asynchronously incoming messages. Hence, we employ a coordinator called the root agent whose responsibility is to collect messages, check the availability of receiving agents, and route each message to the proper agent. Consequently, the communication module of an agent is simplified since the root agent prevent each agent from receiving any message while it is busy performing other task.