<title>Implementation studies of a biologically based controller for a shallow water walking machine</title>

Abstract

Crusteans have been engineered by evolution to adapt and survive in a shallow water hydro- dynamic environment. We have formed a partnership between biology and engineering to explore how crusteans navigate in shallow waters and we are continuing studies to reverse- engineer the biological system. In particular, we are studying the neural control mechanisms of the American lobster, and applying the resultant neurophysiological models to robot control. Our robot is based on a central pattern generation model rather than a reflex chain model of the underlying mechanisms. We have formalized the basic structure of this model in terms of central pattern generators, command, and coordinating systems. The central pattern generator controls the motion of the leg appropriate to walking in all directions with and without sensory feedback. The coordinating system controls the gait pattern of the legs. Proprioceptive reflexes are used to alter the path of a leg due to obstacles and exteroceptive reflexes alter the patterns of several legs to achieve overall motions of the lobster. We have built a simulator which models the central pattern generator and coordination and load compensation. It demonstrates basic walking motions in any direction and at multiple speeds. If additional loads are applied while walking, the neural activity is increased (recruitment) and the leg pattern slows to reflect the additional load. We are currently building an eight legged terrestrial walking machine based on the morphology of the lobster to demonstrate the results of our simulation. In the future we will add more sensors, reflexes, and taxis to our simulator and robot. We plan to implement an underwater version of this robot.