IEEE Robotics & Automation Magazine - December 2017 - 56

mean power output. Finally, during
phase V, a high increase of the maximum power output and the mean poScale Pulsewidth (0-500 µs)
wer output as well as a delay of the
Scale Current (0-120 mA)
fatigue onset was observed, as shown in
ScienceMode
RehaStim
Figure 10. The measurements were
only conducted at weeks one, five, and
Scale Frequency (20-40 Hz)
nine to obtain information on the
training improvements for the final
Emergency Button
race. During these measurements, the
stimulation intensity was set to maxFigure 9. The user interface of the RehaBike. The pulsewidth and current were modified
imum value and kept constant over the
with the same potentiometer, while the frequency was modified with a unique
potentiometer. The potentiometer offers 41 detents, which are distributed linearly over
whole measurement phase.
the full range, preventing unintended slipping control during cycling. The emergency
Prior to each training, our athlete
button was integrated into the right handlebar.
did a 5-min warm-up at a low speed
and low resistance. In summary, the
cadence in phases I-V was successively increased to the same
30
level as during mobile cycling to train our pilot for the conFirst Week
trol interface. Furthermore, during phases I-IV, only quadri25
Fifth Week
ceps and hamstrings were involved. Additional stimulation
Ninth Week
20
of the gluteus maximus only took place in phase V. We tried
15
to avoid the stimulation of gluteus until phase V, because
placing the electrodes at the gluteus was not possible for our
10
athlete without assistance of a third person and would have
5
increased the complexity of the system. However, the gluteus
0
1
2
3
4
5
6
7
8
9 10
stimulation improved the regular cycling performance durTraining Time (min)
ing the dead phases of the two other muscle groups.
Figure 10. The power output for one exemplary 10-min training
During the regular cycling training, a control strategy
session after one, five, and nine weeks of training in phase V. The
was developed by our athlete, which kept the cycling
power output was measured using the displayed power output
of the MOTOmed viva2 ergometer. The compared session was
cadence between 30 and 40 RPM. Below 30 RPM, a small
always the first of the daily sessions.
stimulation pause would disturb the smooth cycling pattern, thus resulting in reduced cycling performance, distance, and speed. Above 40 RPM, a faster fatigue could be
observed. During the first 600 m, our athlete used the gear
and the intensity to keep the cadence in the target area. The
stimulation frequency was increased during the last 150 m
to compensate for fatigue effects and to get the maximum
force for the finishing sprint. In addition to the stationary
ergometer training, we did a regular mobile cycling training
with the RehaBike in the corridors of our institute and,
three weeks before the Cybathlon, at a gymnasium. During
several trials, we sought to find the optimal control strategy
by trial and error. The main goal was to treat the muscles
with care so as to reach the full distance and stay below the
allowed time limit of 8 min. Right before the Cybathlon,
Figure 11. Our athlete during the Cybathlon race. (Photo
our pilot was able to cycle the full race distance of 750 m
courtesy of ETH Zürich, Nicola Pitaro.)
five times with 5-min rests in between.
5-10 W, allowing the pilot's muscles to recover. The high- and
During the preparation, we varied the electrode placelow-power intensity intervals were repeated sequentially until ment and sizes to find the optimal electrode settings for
the daily training target was reached at an energy benchmark our athlete. The final settings were not changed after the
of 150/630 kcal/kJ. At the beginning of phase V, our pilot start of phase V. On the quadriceps, we placed two elecreached this benchmark after 90 min of training, and, right trode pairs (RehaTrode, 5 cm × 9 cm, Hasomed) with two
before the Cybathlon, we reached the benchmark after 50 min. separate stimulation channels in a row. One pair of larger
From phases I to III, we observed only a low increase of mean electrodes (RehaTrode, 7.5 cm × 13 cm) was used for the
power output from 5 to 8 W. Two months after starting daily hamstrings. At the gluteus maximus muscle, one electrode
training in phase IV, there was an increase from 8 to 10 W pair (RehaTrode, 5 cm × 9 cm) was attached.
Power Output (W)

Embedded Control System

IEEE ROBOTICS & AUTOMATION MAGAZINE

DECEMBER 2017

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