Multi-agent systems

Abstract

Structurally, an agent is a bundle of sensors, decision-makers and actuators. Behaviorally, an agent is a mapping from an in-space (the set of things the agent can sense) to an out-space (the set of things the agent can affect). Cells, ants, computer programs, robots and people are examples of agents. Larger agents (multi-agent systems) are organizations of lesser agents. Immune systems, nervous systems, multi-cellular organisms, ecologies, insect societies, distributed computing, communication networks, neural networks, evolutionary algorithms, artificial life, economies, corporations, the Internet, and the control systems of electric grids, are examples of multi-agent systems. This paper presents a key research issue is to find procedures for determining good mixes of cooperation, competition, learning and destruction. Another issue is how to make the other choices involved in designing a multi-agent system.