Anticipatory Networks: Concepts and Selected Applications

Abstract

Anticipatory networks (ANs) generalize and extend the theory of anticipatory systems, which has recently gained popularity in the field of autonomous robotics. Decision analysis in ANs is based on the assumption that the preference model of one decision maker—causally linked to another—takes into account the outcomes of future decision problems solved by other agents. The future decisions relevant from the point of view of the first agent are identified through anticipatory feedback relations. In this lecture, we will present a constructive approach that have been designed and implemented to model and optimize the anticipatory behavior of causally connected decision makers, specifically within teams of autonomous mobile robots. Cooperating robot formations can be modeled as evolving ANs, governed by discrete event systems with virtual supervisors. We will provide an illustrative example involving the determination of an optimal anticipatory decision strategy for a team of four autonomous agents, along with a brief overview of additional applications of ANs.