IEEE Robotics & Automation Magazine - December 2017 - 84

Walker

Parallel Bar

Four-Legged Crutches

One-Legged Crutches

Figure 8. The pilot training for standing up and balancing.

these problems, we proposed the FIW, and the pilot was
able to start walking successfully by himself in May 2016.
In addition, the pilot was able to make a turn by using the
crutches while walking.
After tens of training sessions for walking, the pilot
started accomplishing Cybathlon tasks. The pilot could
walk up to 1.082 km/h with the WalkON Suit through
repeated training sessions prior to Cybathlon 2016. While
training for walking with the WalkON Suit, gait efficiency
and performance were monitored with objectively measured data, such as an oxygen cost. Oxygen cost data are
measurements of the energy expenditure per unit distance of walking, and, thus, they are a good means to
measure how the pilot consumes energy while walking
with a powered exoskeleton. As shown in Figure 9, the
oxygen cost data were collected during level walking by a
Cosmed k4b2 manufactured by COSMED, Inc. Concord,
California. At an early training period, the pilot's oxygen
cost was approximately 30 times higher than that of a normal person without a disability. It was decreased from 100

Oxygen Cost (ml/kg per m)

4
0.198 km/h
(2 June)

3
2

0.271 km/h
(14 June)

1
0

0.695 km/h
(25 September)
0

3
4
Time (min)

Figure 9. The performance measurements during level walking.

IEEE ROBOTICS & AUTOMATION MAGAZINE

DECEMBER 2017

to 25% after training and algorithm optimization, meaning
improved gait efficiency.
4) Sofa: Because standing up and sitting down are repeated in
every training, it took only one day for the pilot to master
the sofa task.
5) Ramp and stairs: The ramps and stairs tasks required much
more time, roughly two months for the pilot to master the
tasks. In the first version of the rules for Cybathlon, it was
not clear whether the pilot could hold the guardrails, and,
thus, the pilot practiced these two challenging tasks without holding the guardrail. However, the rule was modified
to allow holding the guardrail, and then the ramps and
stairs tasks became easier for him.
6) Stones and tilted path: After practicing the stones task for a
month, the pilot could clear the task, but his success rate
was not 100%. In particular, the pilot has relatively shorter
legs compared to the other pilots, and, thus, the distance
between stones was particularly challenging to him. The
pilot had to spend too much time clearing the tilted path
(about 2 min) because it had two ramps without a guardrail and maintaining balance was more difficult than level
walking. Therefore, we decided not to participate in the
tilted path task during Cybathlon 2016.
7) Race simulation: Simulation training for the actual race
began a month before Cybathlon 2016. To create an environment similar to the actual stadium, we installed all of
the tasks in the Sogang University gymnasium in Seoul,
South Korea. In addition, we recruited a crowd to create
the same noise as the actual stadium. In these conditions,
the best time record for the pilot to clear all six tasks during practice was 9 min, 42 s, and the best time record for
the five tasks excluding the tilted path was 7 min, 30 s.
Cybathlon 2016
On 8 October, the Cybathlon was held at the Swiss Arena in
Kloten. In the qualification race, the pilot successfully cleared

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2017