A substantial body of research in neurophysiology suggests that a great deal of locality exists in the control architectures of animal locomotion. In particular, experiments have demonstrated that cats are capable of walking with a wide range of gaits on a treadmill after their spinal cord has been severed to isolate the rear legs from the brain and front legs. This underlines the importance of understanding decentralization in the feedback control of sophisticated bio-inspired legged machines. Dr. Kaveh Akbari Hamed has recently been awarded an NSF-NRI award of $612,000 to conduct his research to proposes innovative decentralization schemes for dynamic quadruped walking that drive joint patterns as a function of a single mechanical variable, referred to as the phasing variable. The phasing variable represents the body’s progression through the gait cycle, replacing the role of time in time-invariant feedback controllers. This concept also has connections to biology, as feedback related to hip motion is essential to controlling the progression of knee and ankle patterns in cat locomotion.
Dr. Akbari Hamed’s proposed optimization algorithms systematically tune the parameters of phase dependent local controllers to produce biologically-inspired decentralization from a rigorous mathematical control perspective. Using this optimization framework, he designs robust stabilizing local controllers for an existing quadruped robot. In particular, the 3D autonomous quadruped robot at SDSU (see Fig. below) provides a unique opportunity to implement his proposed decentralized controllers on subsystems of the robot (e.g., individual legs or rear vs. front legs as in spinal cat experiments) in order to achieve robust walking gaits at variable speeds.