IEEE Robotics & Automation Magazine - December 2017 - 71

Inclusion/Exclusion Criteria
for Cybathlon
We required the pilot to have an
American Spinal Injury Association
Impairment Scale rating of either A
(Complete) or B (Incomplete). Before
each session, the pilot was examined
for preexisting bruises, open sores,
skin lesions, or skin irritations. After
each day of training, the pilot was
examined to determine if the use of
Mina v2 caused any bruising, chafing,
or skin irritations.

(a)

(b)

(c)

(d)

Trajectory Design Results
Table 1 lists the parameters used to
(e)
(f)
(g)
determine the trajectories for the swing
and stance legs using our approach.
These parameters were selected heu- Figure 6. An illustration of ramp ascent (a)-(d) and descent (e)-(g) with powered ankles.
ristically through testing to generate
appropriate trajectories that are executable by the pilot. FacTable 1. A Table of the Parameters Used to Genertors such as feedforward motion, pilot comfort, and pilot staate the Swing- and Stance-Leg Trajectories.
bility are examined throughout the tuning process.
Walking
When walking over flat ground, the desired step length is set
to 0.4 m, with a transfer duration of 0.4 s and a swing duration of 1.0 s, resulting in an average velocity of 1.04 km/h
(0.29 m/s). While this results in the transfer phase making up
a considerably higher percentage of the gait than in able-bodied walking, it provides the pilot adequate time to comfortably and reliably reposition the crutches between steps when
walking continuously. Both transfer and swing durations
could be reduced through more training, as well as increasing
step length. However, these parameters produced a fairly
comfortable gait for the pilot during training. He was able to
maintain this walking speed for a significant duration without
substantial fatigue, often training for over an hour at a time.
Higher walking speeds were attempted, but performance was
not reliable enough for the competition, given the relatively
brief training period.
The resulting trajectories for the transfer and swing leg
walking on flat ground are illustrated in Figure 7, which
shows the average swing and stance leg-joint angles during a
step. The duration has been normalized to the step time. The
shaded gray region represents the transfer phase of the walking gait. The resulting trajectories are smooth. There is a significant toe-off motion during transfer, with the change in the
stance-hip angle moving the pilot forward, which helps the
pilot start walking.
Stepping Stones
The stepping stones task at the Cybathlon consisted of a series
of small wooden platforms that the pilot had to step on sequentially, without allowing his feet to touch the ground around the
platform. This required long steps, typically starting from a

Stepping
Flat Ground Stairs Slopes Stones
Is(m)

0.4

0.29

0.31

$ 0.35, # 0.69

hs(m)

0.1

0.15

0.08

0.1

%l s, b

% ls,f

Swing time (s)

1.0

1.6

1.2

1.8

Transfer time (s) 0.4

1.1

0.6

static position. A large toe-off motion was utilized during transfer to help propel the pilot forward off of his trailing foot and
toward the next stone. We found that being able to do so greatly
assisted our pilot in starting his step to the next stone. This task,
in particular, highlights the benefits of including a powered toeoff motion in the walking gait, as we were one of only two
teams able to successfully complete the task during the competition. Figure 8 illustrates how this large toe-off angle helps the
pilot move his weight toward his stance foot before swing.
Stairs
The stair height for the Cybathlon was 18 cm, and we used a
transfer duration of 1.1 s and swing duration of 1.6 s, resulting
in the joint trajectories shown in Figure 9. While this is substantially slower than what is used in flat walking, it allowed
our pilot to shift his weight forward over the leading foot
before raising it. Unlike other exoskeletons, which typically
only step up with one side, our pilot ascended a stair with
each step, as Figure 5 illustrates, similar to typical ablebodied motion. We believe this was made possible by the use
of powered ankle plantar flexion. Our pilot was able to utilize
DECEMBER 2017

IEEE ROBOTICS & AUTOMATION MAGAZINE

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