To perform different types of tasks, robots are required to adopt various gaits and smoothly transition between them. With current techniques, it is possible to create fullbody robot motions starting from a description of a footfall pattern, as well as highlevel specifications such as the desired walking speed. However, the input footfall pattern needs to be hand-designed, which is both tedious and potentially error-prone[1].

The purpose of this study is to develop an automatic approach to generating locomotion gaits for periodic robot motions as well as footfall patterns for transitions. To address this challenge, we propose a mixed-integer quadratic optimization model that generates stable footfall patterns starting from a set of objectives and constraints. We demonstrate the versatility of this model by generating a variety of gaits and transitions for quadruped, hexapod and dodecapod robots, which are tested both in simulation and on physical platforms.

36 pages

Thesis Committee:
Stelian Coros (Chair)
Nancy Pollard

Frank Pfenning, Head, Computer Science Department
Andrew W. Moore, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu