Dynamic stability in bipedal motion: Adaptive control strategies for balance and controlled motion

Abstract

The exploration of dynamic stability in bidedal machines requires a great deal of knowledge about the science of balancing, both equilibrium and motion. Recent work in robotic legged locomotion has concentrated on systems that require three or more legs on the ground at any given time. This research focuses on adaptive control strategies for a bipedal machine that will allow balance and controlled motion with one leg and, if not walking, on two legs on the ground at any given time. Our approach is to optimize a set of balance and motion profiles through extensive simulation and to validate the profiles on an experimental testbed. Once validated as capable of providing dynamic stability, the adaptive control model uses these profiles as nominal control. The sensory input is then used to modify the nominal control to allow precise control at each sampling period. Simply stated, our control model continuously measures the rate of fall of the biped, and adjusts torques at the knees and hips to constrain this fall to dynamic balance and controlled motion. As should be suspected at this time, our control model is sensor driven and does not require a solution to the Lagrangian equations of motion. The result is a faster, less complex, adaptive control process. Our experimental bipedal testbed currently, and repeatedly, exhibits 25 + stable steps on a flat but slightly varied terrain. Current technology could not provide the kind of actuation and measurements necessary to implement our control model; therefore, our team has developed new low pressure, servoed hydraulic systems and sensory devices. Our most recent experimentation has used parallel computing methods and devices in the C + + programming language on a transputer (parallel computer) based Cogent XTM parallel computing workstation. A new dimension to our research is the translation of our knowledge to manufacturing systems and machines. We are currently investigating how our knowledge of limb coordination and reflex can be applied to the coordination of multiple jointed appendages. In addition, we will explore the use of our positioning and balancing technology in the work cell.