Wearable exoskeletons are powerful solutions that can be applied to reinforce and enhance mobility in able-bodied subjects [1, 2], or to restore lost functions of people with motor problems, such as those resulting from aging [3, 4], neurological disorders as spinal cord injury [5,6,7], or others [8,9,10].
Although these robotic devices are reliable in assisting individuals’
locomotion, researchers still struggle to design smart controllers for
exoskeletons that also support balance when needed. Balance support is
currently a serious demand and an often-heard wish of exoskeletons
stakeholders, who consider this a fundamental and necessary skill [11, 12].
Especially during walking, balance becomes even more challenging, as
recovery reactions to unexpected disturbances are often required to
continue the gait cycle. During dynamic tasks, humans can exploit
different balance recovery strategies, and the selected strategy may
depend not only on the magnitude and direction of perturbation, but also
on the perturbation timing within the gait cycle [13, 14]. Ideally, controllers for exoskeletons should be developed to take into account all these possible reactions.
One
of the main issues of current lower-limb exoskeletons to achieve the
challenge of balance is the insufficiency of human–robot interaction.
This interaction is particularly significant when the prone-to-fall user
still has some residual control. In these situations, cooperative
controllers should be used to support in restoring balance only when
necessary (e.g. onset of a potential fall). This may be known as
“assist-when-needed” approach.
Recent studies with exoskeletons
that developed “assist-when-needed” approaches to support balance were
primarily focused on hip control [15,16,17].
The proposed controllers, provide hip torque to adjust the stepping
location, either by supporting hip abduction-adduction (step-width
adaptation) or hip flexion-extension (step-length adaptation). The
assistance is triggered and modulated when perturbations are detected by
using different feedback signals, such as the hip angle [15], the extrapolated center of mass (XcoM) [16], or the estimated leg force [17].
These approaches are not intended to replace human control, but rather
to augment the user’s balance by providing the required assistance in
synergy with the human wearer just after the onset of an imminent fall.
Although
the hip joint is important for controlling the swing leg and preparing
for foot placement, previous studies provided evidence that also the
ankle joint during stance is crucial in balance maintenance [13, 14, 18].
The torque generated around the ankle acts to decrease the body’s
velocity in the direction of the perturbation. Vlutters et al. [14]
demonstrated that humans modulate the ankle joint torque of the stance
leg as a response to antero-posterior (AP) pelvis perturbations. This
ankle torque modulation scales with the provided perturbation magnitude,
and thereby with subject’s center of mass (COM) kinematics after
perturbation. Using the ankle strategy, subjects were able to
eventually slow down the body movement provoked by the external
disturbance.
Despite the demonstrated importance of the ankle
joint, studies centered on ankle-exoskeleton controllers for assisting
in balance during gait, and their effective evaluation with human users,
are still scarce. Some preliminary approaches specifically designed for
ankle balance support were mainly centered on stance situations. An
example is the work presented in [19], in which the authors demonstrate that standing balance can effectively be supported by a strategy based on the user’s COM kinematics. Another example is the work of Ugurlu et al. [20],
where the authors propose a real-time variable ankle stiffness as a
balance control technique for standing with exoskeletons. Unfortunately
this approach was not tested with the human wearer in the loop. Other
methods that do use cooperative ankle-controllers during human
locomotion did not address their effectiveness in counteracting balance
recovery [21].
Finally, there have also been control approaches based on neuromuscular
models that propose ankle balance assistance during walking with
prosthetic legs [22].
Unfortunately, these models do not demonstrate the ability to generate
cooperative human-like balance responses without specific supplementary
additions [23].
In
this work we have the aim of developing a “simple” bio-inspired control
strategy for ankle-exoskeletons that works in synchrony with the human
and effectively cooperates and assists balance recovery during walking.
In our approach, we first detect disturbances to the COM in
real-time by using human’s kinematics responses. Based on this
detection, we trigger the robotic ankle assistance to recover stability.
The assistance delivered by the controller tries to mimic humans ankle
torque modulation [14],
scaling with body kinematics and distributed proportionally over both
ankles based on the weight supported by the corresponding leg. Our
controller presents high levels of transparency during unperturbed
locomotion [24],
and provides appropriate support in synchrony with the human’s
reaction, ensuring the “assist-when-needed” approach. A specific
advantage of the proposed method is that it does not require specific
subject personalization and thereby it can be easily applied without
time-consuming tuning.
Our main hypothesis is that the developed
ankle-exoskeleton controller is capable of reducing able-bodied users’
effort required to counteract unexpected perturbations during walking
without detriment of their stability. Moreover, we expect the controller
to be reliable in both, detecting the perturbations and providing
assistance that works in sync with the user to eventually help in
recovering balance.
More at link.
Yingzhi Xu
Yue Hu
Shixin Xu
Fengzhi Liu
Ying Gao1,2*