Behaviosites: Manipulation of Multiagent System Behavior through Parasitic Infection

Proceedings of the AAAI Conference on Artificial Intelligence, 21

Volume

Technical Papers

Multiagent Systems

In this paper we present the Behaviosite Paradigm, a new approach to coordination and control of distributed agents in a multiagent system, inspired by biological parasites with behavior manipulation properties. Behaviosites are code modules that "infect" a system, attaching themselves to agents and altering the sensory activity and actions of those agents. These behavioral changes can be used to achieve altered, potentially improved, performance of the overall system; thus, Behaviosites provide a mechanism for distributed control over a distributed system. Behaviosites need to be designed so that they are intimately familiar with the internal workings of the environment and of the agents operating within it. To demonstrate our approach, we use behaviosites to control the behavior of a swarm of simple agents. With a relatively low infection rate, a few behaviosites can engender desired behavior over the swarm as a whole: keeping it in one place, leading it through checkpoints, or moving the swarm from one stable equilibrium to another. We contrast behaviosites as a distributed swarm control mechanism with alternatives, such as the use of group leaders, herders, or social norms.