Institute for Mathematics and Its Applications
Robert Full, University of California Berkeley
During slow terrestrial locomotion, a wide base of support and a low center of mass allow many-legged, sprawled postured animals to be highly statically stable. By contrast, at faster speeds, we found that even these animals use dynamic gaits. They can operate like spring-mass systems during running. We have shown that many extraordinarily diverse morphological solutions appear to be adequate for running. To understand the importance of sprawled postured morphology during running, we made a simple two dimensional dynamic model that focuses on the horizontal plane instead of the vertical. Sprawled posture animals produce substantial lateral forces in the horizontal plane that have been ignored. To test whether this model was robust to perturbations, we changed the momentum of its center of mass during running. We speculated that the model could not recover from a velocity perturbation, since it has an extremely simple control system with no explicit feedback circuitry. The horizontal plane model was extremely stable to velocity perturbations. Dynamic coupling of body orientation, leg moment arms and leg force production produced remarkable self-stabilization. To understand the influence of posture on stability, we varied morphology from an upright posture to a more sprawled posture and increased the amount of the lateral force proportionally. Wider stances, and therefore greater lateral forces, resulted in faster recovery to perturbations. Our two dimensional model tells us that control can reside in the mechanical design of the system and can be simple. The control algorithms can be embedded in the form of the animal itself. Data from climbing, turning and negotiation of irregular terrain show that maneuverability can result from minor adjustments in the feedforward pattern used during straight-ahead locomotion. Three dimensional dynamic models and direct muscle measurements reveal that the musculo-skeletal complexes act as an important determinant of the system's behavior that can't be predicted from the neural signal alone.