Dynamic stability of variable stiffness running

Humans and animals adapt their leg impedance during running for both internal(e.g. loading) and external(e.g. surface) changes. In this paper we examine the relationship between leg stiffness and the speed and stability of dynamic legged locomotion. We utilize a torque-driven reduced-order model of running based on a successful family of running robots to show how optimal clock-driven controllers can interact with variably compliant limbs to adapt to changing operating conditions. We show that the leg stiffness adaptation gives, in general, better results than simply optimizing the gait controller and nearly as good as the co-optimization of controller and leg stiffness.