Compliant limb behavior: Exploiting the basic mechanics of biological legged locomotion for the control of legged systems

Abstract

To walk or run, most legged robots feedback-control predefined joint trajectories. Although this approach works in the laboratory, it hardly succeeds on natural ground, because in such an uncertain and irregular terrain this control requires a sensory precision and an actuator bandwidth that are beyond current technologies. No such requirement is found in biology, yet animals and humans cross natural terrain with compelling ease. In fact, evidence is collecting that they control locomotion differently: rather than enforcing predefined trajectories, biological systems seem largely to exploit their inherent dynamics; hence, to build dexterous legged robots, it may be indispensable to understand these dynamics, and to use the resulting insights for alternative control strategies. Here, we review some of our results from investigating animal and human dynamics that indicate such control alternatives. First, we demonstrate that the basic walking and running dynamics can be explained by compliant limb behavior. Second, we show how compliant limb behavior can be exploited to obtain highly dynamic, mechanically self-stable locomotion on irregular ground using simple feedforward control. Third, we show how a local autonomous feedback control enables compliant legged systems to self-adapt to uncertain terrain. And finally, we present some current efforts to transfer these suggested control strategies to mechatronic systems.