Abstract We are applying a conserved neurophysiological model to the control of behavior in fieldable underwater robots. The model is based on command neurons, coor- dinating neurons, central pattern generators and extero- ceptive reflexes. We discuss implementations using fi- nite state machines as well as electronic neuronal networks. Fig. 1. Biomimetic lobster robot 1. Introduction Biomimetics attempts to capture, in robots, the per- formance advantages that animals enjoy in the natural environment [1]. Legged arthropods such as the lobster have been navigating the ocean bottom in New England for millions of years and are a proven solution to the problems of navigation, sensing and prey capture. Simi- larly, the undulatory swimming of lamprey has enabled them to survive in the open ocean as well as in rivers and streams. Aquatic organisms are only slightly negatively buoyant and as a result must rely on behavioral solutions to adaptation to surge, irregular bottom types, orientation to gravity and complex structured environments. Marine animals have evolved stereotyped reflexes and action patterns that underlie their performance advantages in adapting to the contingencies of the undersea environ- ment [2]. Both lobster and lamprey are important neurophysiological model systems and as a result there is an extensive literature on the organization of their nervous systems and their capabilities for adaptive behavior [3,4]. This background has permitted the development of detailed models of the mechanisms of motor pat- tern generation, sensory integra- tion and sensory-motor integra- tion [5,6]. Similarly, the lamprey has achieved a high level of un- derstanding as an undulatory swimming system [7]. We are developing walking and swimming robots based on the architecture and behavioral set of these animal models in an attempt to capture these perform- ance advantages in engineered vehicles. These robots are based on biomimetic neurotechnology that includes (1) A physical plant that captures the biomechanical fea- tures, (2) Neural circuit based controllers based on the com- mand neuron, coordinating neuron and central pat- tern generator architecture of animal locomotory system, (3) Myomorphic actuators controlled in a similar fashion as muscle, (4) Neuromorphic sensors that code in the same fashion as animal sensors and (5) A behavioral set based on reverse engineering the command sequences that underlie the behavior of the model organism. Fig. 2 The Lamprey- based Robot Architectures for Adaptive Behavior in Biomimetic Underwater Robots Joseph Ayers Dept. of Biology and Marine Science Center Northeastern University East Point, Nahant, MA 01908 and Institute for Non-Linear Science UCSD, La Jolla, CA 92093 In: N. Kato, J. Ayers and H. Morikawa. (2004) "Bio-mechanisms of Swimming and Flying". Springer-Verlag Tokyo.