Meeting Abstract

P1.165  Sunday, Jan. 4  Modeling Swimming Behavior with Perception-Action Feedback Loops in Autonomous Biorobotic Fish STICKLES, E.M.*; SAKHTAH, H.; DOORLY, N; LIEW, C.-W.; ROOT, R.G.; LONG, J.H.; Vassar College; Vassar College; Case Western University; Lafayette College; Lafayette College; Vassar College elstickles@vassar.edu

In order to model the swimming behavior of fish, we use a method borrowed from behavior-based robotics in which sensorimotor (input-output) systems are represented as perception-action feedback loops (PAFLs). PAFLs are modular feedback systems that causally connect sensory information, neural systems, motor output, and environmental interaction. PAFLs can be designed into an agent-environment system in a variety of patterns: solo, in parallel, in series, or in a nested hierarchy. Using a two-layer nested PAFL design, we have built an autonomous fish-like prey robot that forages for light and avoids, when detected, a stalking predator. This robotic simulation is meant to mimic the neural system of fish when Mauthner and related reticulospinal cells override the central pattern generators, used in cruising, in order to create a rapid escape maneuver. Sensory systems model eyes (photoresistors in the low-level PAFL) and a lateral line (IR proximity detectors in the high-level PAFL). Motor output models an undulating tail, capable of turning, with a vertebral column actuated by a single anterior servo motor. With this system, we are able to analyze the connections between sensory input, neural processing, body structure, mechanical properties, and behavior in the context of a predator-prey ecology. This work was supported by NSF DBI-0442269.