IEEE Robotics & Automation Magazine - December 2014 - 59

Angle (°)

Hip

Knee

40

80

30

60

20

15
10

40

5

20

0

10
0

-20

-5

0

-10
0

20

40
60
Stride (%)
(a)

80

100

-20

-10
0

20

40
60
Stride (%)
(b)

80

100

0.8

1

Torque (Nm/kg)

Ankle
20

+

-

0.5

-15

0

-

+

0.4

1.5

0.2

1

-0.5

0

0.5

-0.2

20

40

60

80

100

-0.6

100

40

80

100

+

0

-0.4
0

80

-

0

-1.5

40
60
Stride (%)
(c)

2

0.6

-1

20

0

20

40

60

80

100

-0.5

0

20

60

Stride (%)

Stride (%)

Stride (%)

(d)

(e)

(f)

Baseline

No Support

Support (Total)

n=4

Support (Exotendon)
Figure 4. (a)-(f) Walking kinetics and kinematics. The lines represent the average over the subjects. The average strides start and end
with a heel strike. The top row shows the joint angle for the (a) hip, (b) knee, and (c) ankle joint. For the hip and knee, no significant
changes were found between the conditions. The maximal ankle dorsiflexion angle during support decreased by 5.0 ° compared with
no support and 5.1° compared with the baseline condition. The bottom row shows the joint torque for the (d) hip, (e) knee, and (f)
ankle joint. The shaded areas are the standard deviations.

Metabolic Cost
The metabolic power was calculated from the respiratory
data. The following empirical relation for the metabolic
power ^Eo 6W/kg@h was used [15]:
16.48 $ 10 3 Vo O 2 + 4.48 $ 10 3 Vo CO 2
Eo =
,
m

(1.1)

where Vo O 2 [L/s] and Vo CO 2 [L/s] are the oxygen uptake and
carbon dioxide production, respectively, whereas m [kg] is
the mass of the subject. For all reported metabolic powers, the
metabolic power at rest has been subtracted. The differences
between conditions were compared by the average over the
subjects. A statistical analysis was done by a Wilcoxon signed
rank test.
Results
Kinetics and Kinematics
Figure 4 shows the walking kinetics and kinematics. The exotendons changed the walking kinematics. The maximal ankle
dorsiflexion angle during support decreased by 5 c compared

with the no support condition and 5.1 c compared with the
baseline condition (p < 0.05). The average ankle plantarflexion
torque increased in the support condition by 0.025 Nm/kg
compared with the no support condition and 0.022 Nm/kg
compared with the baseline condition. This increase was not
present for all subjects. The differences in kinematics and kinetics between the conditions were small for the hip and knee.
Differences in stride time between the conditions were also
small (maximal 1.36%) and not significant (p > 0.1).
The measured average absolute human torque was compared with the estimated value from the optimization
(Figure 5). The optimization results predict a decrease in the
average absolute joint torque of 17% for a subject of 70 kg
(this value ranges between 16.8 and 18.3% for subjects between 60 and 100 kg). Experimentally, we found a reduction
of 12.1% in the support condition relative to the no-support
condition (p = 0.089), which could almost entirely be contributed to the ankle torque. The reduction for the ankle only
was 29.0% (p = 0.057). Apart from the differences between
conditions, the human torques in the experiment differ from
the human torques in the optimization. The data for
December 2014

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

59
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2014
