Specifying Agent Behavior as Concurrent Tasks: Defining the Behavior of Social Agents

摘要

Software agents are currently the subject of much research in many interrelated fields. While much of the agent community has concentrated on building exemplar agent systems, defining theories of agent behavior and inter- agent communications, there has been less emphasis on defining the techniques required to build practical agent systems. While many agent researchers refer to tasks performed by roles within a multiagent system, few really define what they mean by tasks. We believe that the definition of tasks is critical in order to completely define what an agent within a multiagent system. Tasks not only define the types of internal processing an agent must do, but also how interactions with other agents relate to those internal processes. In this report, we define concurrent tasks, which specify a single thread of control that defines a task that the agent can perform and integrates inter-agent as well as intra-agent interactions. We typically think of concurrent tasks as defining how a role decides what actions to take, not necessarily what the agent does. This is an important distinction when talking about agents since hard- coding specific behavior may not be the ideal case. Often agents incorporate the concept of plans and planning to determine what to do. In these cases, we would develop a concurrent task for determining how the planning and plan implementation occurs, but not to describe the individual plans themselves.