IEEE Robotics & Automation Magazine - March 2023 - 75

method, the four omniwheels allow the
RRDS to move in any direction, even
in a confined space, and they can also
assist the RRDS in rehabilitation
through walking exercises.
"
THEREFORE, THE CONTROL
TECHNIQUE FOR ENSUR■
The safety seat, as shown in Figure 1(e),
is connected to the force sensor and
is placed under the buttocks of the
rehabilitee; it is used to deal with the
occurrence of unexpected situations.
The rehabilitee who is about to fall
down will sit on the safety seat first
due to insufficient leg strength. At
this point, the safety seat's sensor can
detect pressure, causing the RRDS to
stop moving gradually to keep the robot performing safely
and prevent the rehabilitee from falling.
During the rehabilitation training process, the walking ability
ING THE SAFE MOVEMENT
OF HUMAN-ROBOT
SYSTEMS IS A KEY FACTOR
IN THE DESIGN OF
REHABILITATION ROBOTS.
„
of the rehabilitee is relatively weak at the beginning. Typically, the
rehabilitee must move by mimicking the velocity of the RRDS,
which is passive training. The rehabilitee can walk at the velocity
of his own willingness to engage in the walking movement,
which is active training, as his walking ability improves. To help
the rehabilitee train and develop its intelligence, the RRDS must
now determine the rehabilitee's walking velocity and decide how
to change its own velocity to keep up with the rehabilitee. Regardless
of whether the rehabilitee is receiving passive or active
training, RRDS must undoubtedly follow the training trajectory
recommended by the physiotherapist to realize tracking motion.
Several studies have been conducted on rehabilitation
robots with passive and active hybrid training. However, in
most of the proposed models, the robot needs to be set to stop
the current state of motion when switching the training mode,
and then it should be restarted to enter the next training mode.
However, changing the training modes back and forth by stopping
is neither comfortable nor convenient. Moreover, the legs
of rehabilitees are usually weak, and the affected limbs may
be injured because of the movement inertia of the robot. If a
robot can directly enter the active training stage from the passive
training stage, it can align with the characteristics necessary
for walking rehabilitation training.
CONTROL METHOD FOR PASSIVE AND ACTIVE
DIRECT SWITCHING TRAINING
RRDS DYNAMICS MODEL WITH HUMAN-ROBOT
UNCERTAIN ENVIRONMENTS
The RRDS has four omnidirectional wheels driven by dc
motors. Older people or people with lower-limb disabilities
can be included in rehabilitation training. The structural
coordinates of the robot are shown in Figure 2.
Here
R xC yll
,,
velocity of the robot, and vo is the velocity of each wheel. a is Ox
R xO y^h is the global coordinate system, and
^h is the translation coordinate system. v is the motion
,,
the angle between v and the xl axis, and b is the angle between
r0 and the xl axis. fo
is the control input force for each driven
FIGURE 2. Structure coordinates of the RRDS.
MARCH 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
75
G
r0
C
MM
=
wheel, L is the distance between the center
of the RRDS and each wheel, lo
is the
distance between the center of gravity of
the RRDS and the center of each wheel,
and r0 is the distance between the center
of gravity of the RRDS and the center C,
,,,.
o 1234=
During actual training processes, rehabilitation
robots need to track the paths
specified by doctors. Since the postures
and leg strengths of the rehabilitees are
different, rehabilitation robots encounter
uncertain environments. If the uncertain
environments of a human-robot cooperative
motion cannot be solved, the RRDS
cannot obtain accurate trajectory tracking. Therefore, it is
important to construct a dynamic model with uncertain environments
for rehabilitation robots.
The dynamic model of RRDS is described as follows [17]:
MX tM Xt Bu t12 i+=
po
()
12 and ()B i
()
with the following definitions:
■ ,,
Xt xt yt
i t
() ()
(1)
denote the coefficient matrices, and
■ X(t) is the actual motion trajectory of the RRDS, and
() [( )( )( )] .
their specific meanings are as in [17].
T
■ u(t) is the control input force, and () [] .
ut fff f T
= 1234
■ i is the angle between the horizontal axis and the line connecting
the center of the robot and the first wheel. That is,
.
ii1 =
ii ri ir
=+ =+ and ii4 32 / .r=+
rehabilitee's pose are separated from system model (1); that
is,
As can be observed from the robot structure,
23 / ,,2
The physical quantities affected by the changes in the
MM .Maa1=+D Thus, the RRDS dynamic model with
uncertain environments can be established as follows:
y
ν2 (f2)
y′
l1
β θ
φ
L
ν
α
x′
ν1 (f1)
ν3 (f3)
ν4 (f4)
IEEE Robotics & Automation Magazine - March 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2023
