Rehabilitation technology is at the threshold of a new age when orthotic and prosthetic devices will no longer be separate, lifeless mechanisms, but will instead be intimate extensions of the human body – structurally, neurologically and dynamically. Such a merging of body and machine will not only increase the acceptance of the physically challenged into society, but will also enable individuals suffering from limb dysfunction to more readily accept their new artificial appendages as part of their own body. Several scientific and technological advances will accelerate this mergence, including the development of actuator technologies that behave like muscle, control methodologies that exploit principles of biological movement, and device architectures that resemble the body’s own musculoskeletal design. In this lecture, I first discuss how leg muscles and tendons work mechanically during walking in order to motivate the design of efficient orthotic and prosthetic limbs. I hypothesize that a prosthetic leg comprising only knee and ankle quasi-passive elements, including springs, clutches and variable-damping components, can capture the dominant mechanical behavior of the human knee and ankle during level-ground ambulation. As a preliminary evaluation of this hypothesis, I put forth a simple model that captures the gross features of the human musculoskeletal leg architecture. Model predictions are in good agreement with experimental gait data, suggesting that knee and ankle motors are not necessary for level-ground prosthetic leg ambulation. This result is in support of the idea that muscle-tendon units that span the human knee and ankle mainly operate as tunable springs in walking, affording the relatively high metabolic walking economy of humans. The result also highlights the importance of agonist-antagonist actuation and polyarticular limb architectures in the design of efficient, low-mass, and quiet legged systems for walking. In addition to orthotic and prosthetic biomimetic leg design, the lecture emphasizes the importance of harnessing both zero-moment and moment balance control strategies for the enhancement of bipedal stability and dynamic cosmesis. Recent bipedal walking controllers have included the capability of controlling angular momentum explicitly (Yokoi et al. 2001; Sugihara et al. 2002; Nisiwaki et al. 2003; Popovic et al. 2004; Komura et al. 2005). Many of these controllers tightly regulate angular momentum to be zero at all times. While this is consistent with human behavior during normal walking, such a high level of regulation is not always desirable. For example, when foot placement is constrained, it is often necessary to temporarily sacrifice goals of regulating angular momentum in favor of the more important goal of maintaining balance. In the lecture, I discuss situations where angular momentum should not be zero, and I present a controller that automatically integrates moment-inducing balance strategies with non-moment strategies. The controller uses a multi-variable optimal control approach to automatically balance competing goals for translational and angular momentum. Performance limits of the controller are presented, and compared with those of human test subjects. The lecture concludes with a general discussion of the importance of biologically-inspired hardware and control architectures in the implementation of highly functional legged systems for prosthetic, orthotic and robotic applications.