The dynamics of adaptive behavior: A research program

This paper provides a comprehensive overview of the motivations, methodology and current status of an ongoing research program whose long-term goal is to elucidate the essential principles of a theory of adaptive behavior. The thoroughly dynamical nature of both adaptive behavior itself and the causal mechanisms that support it is emphasized throughout. An initial mapping of the basic concepts of adaptive behavior into the language of dynamical systems theory is proposed, and some of the general consequences of this preliminary theoretical framework are discussed. The two key ideas of this framework are (1) that an agent and its environment should be understood as two coupled dynamical systems whose mutual interaction is jointly responsible for the agent's behavior, and (2) that an agent's need to maintain its existence in its environment defines a viability constraint on its behavioral dynamics. A constructive research methodology involving the use of evolutionary algorithms to evolve continuous-time recurrent neural networks for controlling the behavior of model agents is described, and several examples of this methodology are presented, including models of chemotaxis, walking, sequential decision-making and learning. Finally, a detailed dynamical analysis of one evolved walking circuit is presented. This analysis illustrates the kinds of insights that can be obtained by treating agents as dynamical systems and applying the tools of dynamical systems theory to their behavior.