Autonomous Agent Modeling of Cartilaginous Fish
John Long (Biology and Cognitive Science)
Autonomous agents are lifeforms and any machine that behaves without a human controlling its moment-by-moment actions and reactions. Because complex machines are simpler than even the simplest lifeform, the principles for their behavioral autonomy are well formulated in cognitive robotics. Thus this work on intelligent machines provides a rich set of tools for modeling the autonomous behavior of animals. Beginning with behavior-based subsumption architectures, we delineate behaviors and then coordinate their interactions in order to model the development, behavior, and evolution of cartilaginous fish. Cartilaginous fish -- a taxon that includes sharks, rays, and skates -- are a fascinating group of large marine predators. They navigate, patrol, forage, and attack by integrating an array of sensory systems: olfaction, electroreception, vision, and vibrations. These sensory systems become perceptual systems when coupled with physical actions. Thus perception-action feedback loops are the foundation of our computational work. We build and test models of autonomous goal-directed behavior in developing and evolving cartilaginous fish.