Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network

  Advanced Search  

Human Based Cost from Persistent Homology for Bipedal Walking

Summary: Human Based Cost from Persistent
Homology for Bipedal Walking
Ram Vasudevan*, Aaron D. Ames**, and Ruzena Bajcsy*
* Electrical Engineering and Computer Sciences ** Mechanical Engineering
University of California, Berkeley Texas A&M University
Berkeley, CA 94720 College Station, TX 77843
{ramv,bajcsy}@eecs.berkeley.edu aames@tamu.edu
Abstract: While the focus of robotic bipedal walking to date has been the development of
anthropomorphic gait, the community has been unable to agree on a model for such gait. In this
paper, we propose a universal ordering of events for bipedal walking based on motion capture
data collected from a walking experiment. We process the motion capture data using persistent
homology to automatically determine the ordering of discrete events. Surprisingly, every subject
in the experiment had an identical ordering of such events. This universal ordering allows us to
propose a cost function based upon human data: the human-based cost.
The goal of bipedal walking is typically not to minimize
a concrete cost, like torque squared or the specific cost of
transport, but rather to achieve the more ambitious goal of
obtaining human-like walking. The design of a controller to
achieve this objective is implicitly related to the discrete


Source: Ames, Aaron - Department of Mechanical Engineering, Texas A&M University


Collections: Engineering