IEEE Robotics & Automation Magazine - December 2015 - 88

Figure 7. The gait cycle starts with a heel strike, and only the
CF portion of stance when the damping control works is
shown. The reference angle data are from [20] and [21].
According to the figure, a prosthesis in the maximal damping
mode without control can only move within a limited angle
range (less than 8c). The ankle angle of the prosthesis without
control starts with 0c for
all the terrains, which is
much different than an
The braking torque was
intact limb.
The proposed robotic
estimated by the average
prosthesis has an angle
range of !25c, which is
armature current, with the
similar to that of the intact limb. With the protorque constant of
posed damping control
strategy, the robotic pros33.5 mNm/A provided by
thesis can adjust the ankle
angle during swing and
the motor data sheet.
prepare for the new terrain of the next step.
During stance, the damping of the ankle is adjusted according to the ankle angle, and the resulted angle trajectories can
effectively mimic those of the intact limb, especially on terrains such as LG, SA, and RA. Angle trajectories during SD
and RD are not quite similar to those of the intact limb. For
the initial foot strike with the ground during SD, the amputees could not manage the toe strike with a plantar flexion
angle as large as able-bodied individuals, so the initial angle
of SD was set to be a smaller plantar flexion angle (6c). For
RD, the CP was much slower than that of the intact limb,
causing the trajectories to be quite different. This discrepancy was due to the fact that the amputees tended to be more
careful when the prosthetic foot contacted the ground.
Gait Indicators
The calculated indicators of the proposed robotic prosthesis
on different terrains are shown in Figure 8. The CoP shift in
ML with the robotic prosthesis had a significant decrease
from the passive one during SD and RD, while the indicator
values on the other terrains did not change much. As for the
CoP shift in AP, the robotic prosthesis had similar values to
the passive one on LG and stairs, while the values on ramps
had a significant increase, because the robotic prosthesis with
damping control enabled full contact between the prosthetic
foot and the ramp. Normalized stance time with the robotic
prosthesis had greater values than that with the passive prosthesis, which indicated that the robotic prosthesis created a
more symmetrical gait. As for the normalized peak GRF, values with different prostheses during SA did not vary a lot,
while those with the robotic prosthesis were smaller than values with the passive prosthesis, especially during ramp ambulation. Smaller peak GRF implied a smaller impact on the
residual limb. The four indicators show that the robotic prosthesis has improved the amputee's gait symmetry and walking stability.
88

IEEE ROBOTICS & AUTOMATION MAGAZINE

DECEMBER 2015

Discussion
For the emphasis on terrain adaptation, instead of providing
large assistive torque, the proposed robotic prosthesis with a
low-power motor is quite lightweight. With the proposed damping control method, the prosthesis can manipulate the ankle impedance during stance and enable the amputee to smoothly
move on different terrains. The amputee's gait symmetry and
walking stability are improved, and the prosthesis with damping
control has similar angle trajectories to the intact limb on LG
and ascent terrains. The terrain adaptation ability of the prosthesis does improve the amputees' bipedal walking performance.
Compared with Ossur's Proprio Foot, which can only adjust the ankle angle during swing and lock the ankle during
stance, allowing the carbon-fiber foot to store significant energy as it deflects in dorsiflexion and return it in plantar flexion [15], the PKU-RoboTPro employs a different strategy,
trading the ability to store and return significant energy in a
carbon-fiber spring for the capability of adjusting the ankle
impedance during stance and adapting to different walking
speeds and terrains.
Due to the lack of a parallel spring as well as a high-power
motor, the PKU-RoboTPro cannot provide high assistive
torque like the powered prostheses developed by BiOM [12]
or SpringActive [13]. However, the lighter weight of the prosthesis can reduce the load of the residual limb and help the
wearer to move more easily during the swing phase. In addition, the prosthesis consumes less energy during one gait
cycle, which means the walking distance is farther for the
same battery capacity.
Other reported active prostheses, e.g., the AMP-foot prosthesis, can output sufficient power during the late stance
phase, but the ankle is not controllable during the early stance
phase and the swing phase. Compared with AMP-foot, the
PKU-RoboTPro can not only control the ankle damping during the stance phase but also adjust the ankle angle during the
swing phase to bring foot clearance from the ground and prepare for the next stance phase.
The proposed robotic transtibial prosthesis PKURoboTPro, though promising, has its limitations and needs
further improvements. The main limitation of the proposed
prosthesis is that it can neither provide large assistive torque
nor store significant energy in the carbon-fiber foot to return
to the user. The peak torque of the prosthesis is around 50 Nm,
which is not enough to walk fast. Further optimization for the
balance between joint torque and prosthesis weight is needed.
In addition, though the current system can perform smooth
locomotion on different terrains with ankle damping behaviors, the damping control strategy needs to be improved to enable the prosthesis to mimic the intact limb more effectively.
Further studies on integration of the locomotion mode recognition system and evaluation experiments on more distinguished terrain transitions may bring better gait behaviors.
Conclusions
In this article, we proposed a lightweight robotic transtibial
prosthesis with damping behaviors for terrain adaptation.

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2015