Robot Locomotion Controller Generation Through Human-Inspired Optimization
MetadataShow full item record
This thesis presents an approach to the formal design, optimization and implementation of bipedal robotic walking controllers, with experimental application on two biped platforms. Standard rigid-body modeling is used to construct a hybrid sys- tem model of robotic walking; this model estimates the motion of the robot hardware under a given control action. The primary objective of this thesis is the construction of a control law which effects, on the robot, a periodic “walking” behavior. The pro- cess begins with examination of human walking data—specifically outputs of human walking—which provide inspiration for the construction of formal walking control laws. These controllers drive the robot to a low-dimensional representation, termed the partial hybrid zero dynamics, which is shaped by the parameters of the outputs describing the human output data. The main result of this paper is an optimization problem that produces a low-dimensional representation that “best” fits the human data while simultaneously enforcing constraints that ensure a stable periodic orbit and constraints which model the physical limitations of the robot hardware. This formal result is demonstrated through simulation and utilized to obtain 3D walking experimentally with an Aldebaran NAO robot and NASA’s prototype Leg Testbed robot.
Powell, Matthew Joseph (2013). Robot Locomotion Controller Generation Through Human-Inspired Optimization. Master's thesis, Texas A & M University. Available electronically from