IEEE Robotics & Automation Magazine - June 2013 - 72

dynamic walking. They make use of gravity to make their
underactuated robots walk down inclined slopes without
being controlled and without energy consumption [6]-[8].
Researchers have been interested in building robots that can
walk on a flat surface with minimal energy consumption.
They use simple control methods and a minimal number of
motors to make their robots walk [9]-[11].
To make more humanlike robots, an increasing number of
research groups have developed humanoid robots with spinal
joints. Mizuuchi et al. built a tendon-driven robot called
Kenta [12], [13]. Although the robot has a spinelike structure,
only photos of the robot making static postures while sitting
with external support have been shown. There have been no
data to show that it can move its spine autonomously in a
flexible way. Also, the robot cannot stand up without external
support because of the heavy upper body [13]. Mizuuchi et al.
later developed the Kotaro. Although the robot can neither
stand up nor walk without any external wires supporting it
from above, it can bend its upper torso laterally while sitting
on a chair [14]. However, such movement could be achieved
by other robots if a roll-axis joint was added at the waist. A
later version of Kotaro was made with reinforced muscles.
Again, no experimental data are available to show that it can

stand up [14]. Mizuuchi et al.'s latest prototype is named
Kojiro. It cannot stand up without support while moving on
the spot [15], [16]. Based on the tendon-driven approach,
Holland and Knight developed an upper-torso robot called
CRONOS [17]. Billard and coworkers developed a bendable
upper-torso robot that is fixed on a platform with a hydraulic
pump [18]. Sugano and coworker developed the TwendyOne, which has a 4-DoF spine but is fixed on a wheeled
mobile platform [19]. Similarly, Hirzinger and coworkers
developed a wheeled robot called Justin that has a 3-DoFarticulated torso [20].
Given that it is very difficult to control a flexible spine
humanoid robot, we sought inspiration from belly dancing
and found that a lot of complex torso movements exhibited
by belly dancers are rhythmic and wavelike [21]. We then
developed the world's first full-body humanoid robot
(WBD-1) that is capable of bending its upper torso dynamically while standing on the spot without external support
[22]. The robot's rhythmic and wavelike spinal motions were
controlled using a model of the central pattern generator
(CPG) of a lamprey. Later, we developed a robot named
WBD-2. This robot is capable of expressing emotions using
belly dance movements [23], [24]. Inspired by the way fashion
models walk, we developed the
world's first humanoid robot
(WBD-3) that is capable of walking
with spinal motion. Unlike a conHead
ventional humanoid robot, our
322 mm
robot can walk like a woman with
221.5 mm
swaying hips [25].
Neck RPY
By combining ideas from biology
281
mm
Shoulder (RSH) RPY
RPY
and engineering, we developed a
348.5 mm
hybrid CPG-zero moment point
260 mm
SEG3
RPY
(ZMP) control system to allow a
106 mm
four-segment (SEG1-4) mechanical
SEG2
RPY
Elbow (REL)
spine to maintain balance in real
P
P
106 mm
time while exhibiting belly dance
SEG1
RPY
220 mm
movement [26]. We then developed
212 mm
the world's first control system to
90 mm
Hip (LTH)
Wrist (RWR) RPY
allow a realistically simulated fullRPY
RPY
RPY
RPY
body spine robot to maintain balWaist (WST)
ance in real time while standing
Hand
300 mm
[27]. Later, we developed the world's
Thigh
first control system to allow a simulated robot to maintain balance in
Knee (LKN) P
P
real time while performing dynamic
walking with spinal motion [28].
270 mm
Shin
Although there have been a few
studies using the lower body to
obtain efficient bipedal walking, the
RPY
Ankle (LAK) RPY
contributions of the spine on energy
135.5 mm
Foot
consumption have not been investigated. In this article, we present a
systematic and generalized techFigure 1. Dimensions of the simulated humanoid robot. R, P, and Y correspond to the
nique to find the styles of spinal
roll-, pitch-, and yaw-axis joints, respectively. The dot in purple represents the COM.
motion that lead to efficient walking.
For configurations of the robot, see [33, Figure 1].
72

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

june 2013
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2013
