Skip to main content
eScholarship
Open Access Publications from the University of California

UCLA

UCLA Electronic Theses and Dissertations bannerUCLA

Development and Real-Time Optimization-based Control of a Full-sized Humanoid for Dynamic Walking and Running

Abstract

Creating machines that resemble a human morphology have always been an aspiring goal for mankind even from the early days of engineering. Today, this effort continues to exist especially because of how much adaptive and flexible humans can be, as we can carry out a wide variety of tasks by collectively using our arms and legs. We can run, fall, get back up, and even use our arms or tools to provide additional stability when carrying out different locomotion or manipulation tasks. Imagine how much more helpful machines could be in our lives if we could build human-like machines that could do even a fraction of what we can. Traditionally however, humanoids have been big, bulky machines that moved around very slowly and were susceptible to disturbances and falling down.

Recently, with the collective advancement of key technologies, we have seen a rapid growth in the number of quadrupeds that can now dynamically walk and run in outdoor environments. Quadrupeds are inherently a stable platform, which is opposite to a biped that is inherently unstable. Yet, the underlying core principles in both hardware and software could be selectively adopted in a humanoid context to enhance their agility and robustness.

Hence, this dissertation is an extension of such ideas on a humanoid to make our human-like machines dynamically walk and run such that they can do more meaningful tasks in the future. This requires a development of a hardware platform that uses similar design principles that were successful on quadrupeds, as well as a software and control stack that uses state-of-the-art techniques to robustly control the robot. Therefore, this dissertation introduces ARTEMIS, the first full-sized humanoid robot using proprioceptive actuators, its key design features, along with a real-time optimization-based dynamic locomotion stack that allows ARTEMIS to walk and run. Such hardware and software development allowed ARTEMIS to be the fastest walking humanoid reaching up to 2.1 m/s walking speed, be able to take aggressive pushes from all facets of its body, robustly walk without perception on debris cluttered terrain, and also be the very first humanoid fully developed in academia that can run. This opens up an exciting new chapter in the journey to developing humanoids that not only look like us, but can also robustly move and accomplish meaningful tasks.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View