Effects of upper body parameters on biped walking efficiency studied by dynamic optimization

Kang An,Chengju Liu,Chuanjiang Li,Zuhua Fang

doi:10.1177/1729881416682702

Kang An, Chengju Liu + Show 2 more

Open Access

https://doi.org/10.1177/1729881416682702

Copy DOI

Abstract

Walking efficiency is one of the considerations for designing biped robots. This article uses the dynamic optimization method to study the effects of upper body parameters, including upper body length and mass, on walking efficiency. Two minimal actuations, hip joint torque and push-off impulse, are used in the walking model, and minimal constraints are set in a free search using the dynamic optimization. Results show that there is an optimal solution of upper body length for the efficient walking within a range of walking speed and step length. For short step length, walking with a lighter upper body mass is found to be more efficient and vice versa. It is also found that for higher speed locomotion, the increase of the upper body length and mass can make the walking gait optimal rather than other kind of gaits. In addition, the typical strategy of an optimal walking gait is that just actuating the swing leg at the beginning of the step.

Highlights

Biped walking has become a highlight in robot research in recent years because of the resemblance to human beings
We can find that with the increase of the upper body length, the cost of transport (COT) is increasing when walking with short step length, while the COT is decreasing and getting almost flat when walking with long step length
For the medium size of step length (Dm2 [0.6, 0.8]), the COT curve is almost flat with the increase of the upper body length, which indicates that the effect of lb on the COT is not obvious when walking within this range of step length Dm

Summary

Introduction

Biped walking has become a highlight in robot research in recent years because of the resemblance to human beings. The periodic motion requires that the initial conditions at the beginning of the walking cycle qðt 1⁄4 0Þ must be identical with the step; the time that the swing leg contacts with the ground is at the heel-strike ðt 1⁄4 t stepÞ; all mass points of the model must be above the ground; the velocity of the swing foot along the y-axis just before the heel-strike must be smaller than zero, which ensures that the swing foot lands from above; and the normal force uN at the stance foot should be bigger than zero to satisfy the physical constraint of walking qðt 1⁄4 2q1ðt.

Results and discussions

X v max d max COT vd

Conclusion