IEEE Robotics & Automation Magazine - December 2017 - 55

Important data and parameters could be transferred during
the cycling via a user datagram protocol to an Android tablet
that ran a customized Kivy app (https://kivy.org) for parameter adjustments, data visualization, and logging. All of the
presented methods were implemented on the embedded
control system and were used during our pilot's daily ergometer training and during the mobile cycling exercises.
A detailed risk analysis of the overall system, including a
device manual and available certificates, was approved by the
Cybathlon technical committee according to ISO 13482:2014
[13]. Furthermore, a certified confirmation by our medical
team partner, Dr. Andreas Niedeggen, proved that there were
no objections to our pilot's participation at the Cybathlon
from a medical standpoint. Both technical and medical
checks were accomplished with an on-site evaluation of our
device and pilot at the Cybathlon.
Training and Cybathlon Results
The preparation for the Cybathlon can be divided into stationary training and regular mobile cycling training. The
home training consisted of five phases, which are summarized in Table 4. In the studied literature, no training
regimes for long-term atrophied muscles are described with
respect to FES cycling. Our athlete had been paralyzed for
more than 30 years at the time of the Cybathlon and had no
prior experience with FES. No FES-induced isometric contraction exercises were applied prior to the first cycling
training sessions. According to the ASIA Impairment Scale,
his lesion is classified as Grade A, which means a complete
lack of motor and sensory function below the level of injury. The goal of phase I was to restore the full range of
motion of the lower limbs, to familiarize the skin with the
electrodes, and to reactivate the atrophied muscles. For the
training, the RehaStim stimulator in combination with the
MOTOmed viva2 ergometer (RECK-Technik GmbH & Co.
KG, Betzenweiler, Germany) was used during phases I-III.
The MOTOmed viva2 ergometer offers different operation
modes, such as passive (motor-driving) training, motorassisted training, and training without motor support (the
motor acts as a brake with adjustable resistance). The
power actively produced by the muscles is determined by
the device and shown to the user. It has large foot shells
where the patient places his or her feet. The ergometer
parameters we adjusted during the training were the resistance level and the minimal cadence maintained by the

Original Pulleys

(a)
Oval 23D
Chain Ring

Standard 24/22D Chain Rings of
Eight-Speed Shimano Nexus

Circular 22D Chain
Chain Tensioner
(b)
Figure 8. (a) The original chain configuration including the
pulleys. (b) A schematic of gearwheels and chain of the modified
TRIX drive. (Images courtesy of HASE BIKES.)

motor in case of absent muscle power. To minimize the difference between training and regular cycling, the ergometer
training was done with the MOTOmed viva2 ergometer in
combination with the inertial-sensor-based method in
phase V. Furthermore, our athlete trained during this phase
on the usage of the control interface while manipulating the
load settings of the ergometer (see Figure 9).
During phase I, we used the lowest resistance level and a
minimum cadence of 25 RPM to avoid injuries to the joints
and bones of our athlete. In phase II, the musculature was cautiously built up. The ergometer was used again with the lowest
resistance level, while the minimum cadence was increased to
40 RPM. Phases III and IV were continuations of phase II,
with the minimum cadence set to 30 RPM. Finally, in phase V,
the maximal power increase was intended. We set the minimum
cadence to 35 RPM, the resistance to the highest level, and
modulated the stimulation intensity in intervals. After a short
warm-up phase, we increased the pulsewidth and current until
we achieved a power output of 20-30 W. To compensate for
muscle fatigue, the frequency, current, and pulsewidth were
increased to hold the desired power output for as long as possible. After these high-power intervals, we set the intensity for
5-10 min to low values to have a reduced power output of

Table 4. The preparation phases for the Cybathlon.
Phases

Period

Frequency

Duration

Stimulation Settings

Phase I

April-July 2015

Once a week

30 min

220 ns, 65 mA, 30 Hz

Phase II

August-November 2015

Twice a week

30 min

250 ns, 65-80 mA, 30 Hz

Phase III

December 2015-April 2016

Every second day

30 min

280 ns, 80-90 mA, 30 Hz

Phase IV

May-July 2016

Daily

30 min

280 ns, 80-90 mA, 30 Hz

Phase V

August-October 2016

Twice a day

50-90 min

220-500 ns, 65-120 mA, 25 Hz

DECEMBER 2017

IEEE ROBOTICS & AUTOMATION MAGAZINE

https://www.kivy.org

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