Hierarchical controller for highly dynamic locomotion utilizing pattern modulation and implementation on the MIT Cheetah robot

This thesis presents a hierarchical control algorithm for quadrupedal locomotion. We address three challenges in developing a controller for high-speed running: locomotion stability, control of ground reaction force, and coordination of four limbs. To tackle these challenges, the proposed algorithm employs three strategies. Leg impedance control provides programmable virtual compliance of each leg which achieve selfstability in locomotion. The four legs exert forces to the ground using equilibriumpoint hypothesis. A gait pattern modulator imposes a desired footfall sequence. The control algorithm is verified in a dynamic simulator constructed using MATLAB and then in the subsequent experiments on the MIT Cheetah robot. The experiments on the MIT Cheetah robot demonstrates high speed trot running reaching up to the speed of 6 m/s on a treadmill. This speed corresponds to a Froude number (Fr = 7.34), which is comparatively higher than other existing quadrupedal robots. Thesis Supervisor: Sangbae Kim Title: Esther & Harold E. Edgerton Assistant Professor of Mechanical Engineering