Reconfigurable networking for coordinated multi-agent sensing and communications.

Abstract

An implementation of a neurally-inspired system comprised of multiple mobile sensor-effector agents is described. Each agent has features of a complex neural network that is able to communicate and adjust its behavior depending upon a variety of parameters, including changes in the environment and the behavior of other agents. The system as a whole spatiotemporally reconfigures itself to perform coordinated behaviors not obtainable with single agents. Transient clustering of agents into functional subsystems to perform specific tasks generates a "system of systems" architecture. The interesting findings of this dynamic platform show that (a) the formation and dissolution of functional subsystems is a local phenomenon without the need for global control and (b) minimal intermittent communication among the agents can yield large-scale, coordinated, goal-driven behavior under a wide range of conditions.