SS symmetry

Article
CardioVR-ReTone—Robotic Exoskeleton for Upper Limb
Rehabilitation following Open Heart Surgery: Design,
Modelling, and Control
Bogdan Mocan 1, * , Claudiu Schonstein 2 , Calin Neamtu 1 , Mircea Murar 1 , Mircea Fulea 1 , Radu Comes 1
and Mihaela Mocan 3

1 Department of Design Engineering and Robotics, Technical University of Cluj-Napoca,

400020 Cluj-Napoca, Romania; calin.neamtu@muri.utcluj.ro (C.N.); mircea.murar@muri.utcluj.ro (M.M.);
mircea.fulea@staff.utcluj.ro (M.F.); radu.comes@muri.utcluj.ro (R.C.)
2 Department of Mechanical Systems Engineering, Technical University of Cluj-Napoca,
400020 Cluj-Napoca, Romania; claudiu.schonstein@mep.utcluj.ro
3 Department of Internal Medicine, University of Medicine and Pharmacy Iuliu Hatieganu Cluj-Napoca,
400012 Cluj-Napoca, Romania; mihaela.mocan@gmail.com
* Correspondence: bogdan.mocan@muri.utcluj.ro

Abstract: Following cardiac surgery, patients experience difficulties with the rehabilitation process,
often finding it difficult, and therefore lack the motivation for rehabilitation activities. As the number
of people aged 65 and over will rise by 207 percent globally by 2050, the need for cardiac rehabilitation
will significantly increase, as this is the main population to experience heart problems. To address this




challenge, this paper proposes a new robotic exoskeleton concept with 12 DoFs (6 DoFs on each arm),

 with a symmetrical structure for the upper limbs, to be used in the early rehabilitation of cardiac
Citation: Mocan, B.; Schonstein, C.; patients after open-heart surgery. The electromechanical design (geometric, kinematic, and dynamic
Neamtu, C.; Murar, M.; Fulea, M.; model), the control architecture, and the VR-based operating module of the robotic exoskeleton are
Comes, R.; Mocan, M. presented. To solve the problem of the high degree of complexity regarding the CardioVR-ReTone
CardioVR-ReTone—Robotic kinematic and dynamic model, the iterative algorithm, kinetic energy, and generalized forces were
Exoskeleton for Upper Limb
used. The results serve as a complete model of the exoskeleton, from a kinematic and dynamic
Rehabilitation following Open Heart
point of view as well as to the selection of the electric motors, control system, and VR motivation
Surgery: Design, Modelling, and
model. The validation of the concept was achieved by evaluating the exoskeleton structure from an
Control. Symmetry 2022, 14, 81.
https://doi.org/10.3390/
ergonomic point of view, emphasizing the movements that will be part of the cardiac rehabilitation.
sym14010081
Keywords: exoskeleton; geometric; kinematic; and dynamic model; virtual reality; cardiac rehabilitation
Academic Editor: Jan Awrejcewicz

Received: 24 November 2021

Accepted: 27 December 2021
Published: 5 January 2022 1. Introduction
Publisher’s Note: MDPI stays neutral
Providing structured exercise and education to cardiac patients in an organized man-
with regard to jurisdictional claims in
ner, determining medical risk reduction, means cardiac rehabilitation (CR). CR reduces
published maps and institutional affil- mortality by up to 25%, increases functional capability, and reduces re-hospitalization,
iations. according to strong evidence [1]. A proper rehabilitation program is required to enable
cardiac patients to live independently and improve their quality of life (QoL) as they age [2].
To increase joint range of motion (ROM), reinforce muscles, restore cardiac functional capa-
bilities, and resolve deficiencies, most of these rehabilitation treatments require sessions of
Copyright: © 2022 by the authors. rehabilitation therapy [3].
Licensee MDPI, Basel, Switzerland. Integrating human and robotic-machine capabilities into a single system opens up a
This article is an open access article vast range of new possibilities for assistive technology development. Members of both the
distributed under the terms and healthy and cardiac populations may benefit from its potential uses. Muscle strength is a
conditions of the Creative Commons
limiting factor for many physical undertakings. Muscle weakness is also the leading cause
Attribution (CC BY) license (https://
of impairment in people with a range of cardiac diseases (e.g., those requiring open-heart
creativecommons.org/licenses/by/
surgery) or neuromuscular diseases [4,5]. Even after restoring strength and function to the
4.0/).

Symmetry 2022, 14, 81. https://doi.org/10.3390/sym14010081 https://www.mdpi.com/journal/symmetry

Symmetry 2022, 14, 81 2 of 30

upper limbs, many patients still have weight-bearing asymmetry. Long-term asymmetry
can cause secondary muscle weakening in the upper limbs and even in the lower body [5].
An asymmetry of 10% or more is regularly cited in the literature as an indicator of major
injury risk [4]. Humans have highly specialized and complicated “biological algorithms”
for movement control that involve both higher and lower neural centers. These biological
algorithms allow people to do extremely complex activities, like arm movement, while
avoiding object collisions [6]. Robotic manipulators, on the other hand, can be constructed
to do jobs requiring significant forces or movements, depending on their structure and
actuator power. However, the control algorithms that govern their dynamics lack the
needed flexibility while maintaining the same level of quality as human limbs [7].
As a result, it appears that integrating these two entities—the human and the robot—
into a single integrated system controlled by a human could result in a solution that
outweighs the disadvantages of each subsystem. When the machine’s mechanical power is
combined with the innate human control system, it may be possible to perform activities
that require higher forces than a human could otherwise produce. Two key scientific
and technological concerns lie at the heart of this human–machine integration: (i) the
exoskeleton (orthotic device) mechanism itself and its biomechanical integration with the
human body, and (ii) the human–machine interface (HMI) [8,9].
As an assistive device, the exoskeleton is an external actuated structure with human-
like joints and connections. The exoskeleton is worn by the patient, and its actuators
produce torques that are applied to the human body segments. The human provides
control signals for the exoskeleton, while the exoskeleton actuators provide much of the
power required for task performance. Compared to the weight borne by the exoskeleton,
the human becomes a component of the system and applies a scaled-down force.
The use of an exoskeleton in CR augmented with a VR system is a relatively recent
technique aimed at allowing patients to interact with virtual items in virtual reality. As a
result, the operator can interact with virtual objects [10]. Developing such a mechatronic
system suitable for CR requires comprehensive documentation, considering models of
upper limb exoskeletons used for rehabilitation focused on mechanisms, actuators, sensors,
and control techniques. A series of studies and research regarding the various existing
upper limb robotic exoskeletons are highlighted in Table 1.

Table 1. Selection of existing upper limb exoskeletons.

Control Clinical
Reference/Year Rehab Type Movements DoFs Actuator Type
Type/Feedback Tests
Rosen et al. EMG DC servo and
Active assist Shoulder, elbow 1A + 1P No
[11]/2001 optical incr. encoder gearbox
Fanin et al. Passive or Wire driven by BLDC
Shoulder, elbow 3A Incremental encoder Yes
[12]/2003 active assist motor
DC motor and HD
Nef et al. Passive or Position and force
Shoulder, elbow 4A + 6P gearbox, cable drive Yes
[13]/2007 active assist sensors
linear module
Ball–screw and cable
Garrec et al. Passive or
Shoulder, elbow 4A Force sensor driven by electric No
[14]/2008 active assist
motor
Brackbill et al. Cable system driven
Active assist Shoulder, elbow 4A Encoder No
[15]/2009 by BLDC motor
DC motor driving a
Frisoli et al. Passive or Shoulder, elbow, Force feedback, VR
4A + 1P system of pulleys Yes
[16]/2009 active assist wrist optical encoder
and ball bearings
Stienen et al. Electro-hydraulic
Passive assist Shoulder, elbow 4A + 3P Load sensor No
[17]/2010 disk brakes
Symmetry 2022, 14, 81 3 of 30

Table 1. Cont.

Control Clinical
Reference/Year Rehab Type Movements DoFs Actuator Type
Type/Feedback Tests
Naidu et al. Shoulder, elbow,
Passive 4A + 3P Joint angle Electric actuators No
[18]/2012 wrist
Pirondini et al. Passive or Shoulder, elbow,
4A + 2P EMG, force control No info provided No
[19]/2014 active assist wrist
Zhou et al.
Passive assist Shoulder, elbow 4 No info provided Not applicable No
[20]/2015
EMGs, force sensor,
Wu et al. Shoulder, elbow, Cable system driven
Active assist 7A + 2P potentiometer, virtual Yes
[8]/2018 forearm, wrist by AC servo motor
reality
Islam et al. Active or Shoulder, elbow, Force sensors, Electric motor
7A + 2P No
[21]/2021 passive assist forearm, wrist Hall sensor encoders coupled to HD gear
Schabron et al. Hand gesture control,
Active assist Shoulder, elbow 3A Stepper motor No
[6]/2021 EMGs
Gull et al. Trajectory tracking, BLDC motor coupled
Active assist Shoulder, elbow 3A + 1P No
[22]/2021 encoders to HD gear
Huamanchahua et al.
Passive assist Shoulder, elbow 4A + 1P Virtual reality tracking No motors No
[23]/2021
Galafaro et al. Active or
Shoulder, elbow 4A + 2P Impedance control BLDC motors No
[9]/2021 passive assist
BLDC with belt
Liu et al. Force sensor, EMGs,
Active assist Shoulder, elbow 2A + 4P pulley and gear No
[10]/2020 joint angle
reducer
Chen et al. Position control, joint Pneumatic muscle
Passive assist Shoulder, elbow 3A + 1P No
[24]/2020 angle, encoder actuator
Curz et al. Shoulder, elbow, Linear and rotary
Passive assist 7A PLC contro No
[4]/2020 wrist electric motors
Stereo vision, auditory
Zhao et al. Active or Shoulder, elbow, sensor, EMG, force Cable driven by DC
3A No
[25]/2020 passive assist wrist sensor, proximity, servo motors
encoder
Cable with gear
Buongiorno et al. Active or Shoulder, elbow, EMG, autoencoder,
4A + 1P driven by electric No
[26]/2020 passive assist wrist force sensor
motor
Electroencephalographic
Badesa et al. Shoulder, elbow, (EEG),
Active assist 4A + 8P No data No
[27]/2019 hand, wrist electrooculographic
(EOG)
Chen et al. Encoders, inertial Cable driven by DC
Active assist Shoulder, elbow 4A No
[28]/2019 measurements unit electric motors
Passive or
Chaparro et al. DC motor and
active Shoulder, elbow 2A Encoder No
[29]/2020 planetary gearhead
assistance

By analyzing the solutions and prototypes of the upper limb robotic exoskeletons
developed over time by various research groups and companies, as highlighted in Table 1,
the following conclusions can be drawn:
a. Actuator type: the actuator types can be DC servo, electro-hydraulic disk brakes; AC
servo motor; stepper motor; BLDC motors; or cable system driven by BLDC motor,
with harmonic or planetary gearboxes.
Symmetry 2022, 14, 81 4 of 30

b. Control type/Feedback: the control type and feedback include incremental encoder;
position and force sensors; EMG optical incr. encoder; electroencephalographic
(EEG); and electrooculographic (EOG).
c. Movements: the parts of the body that the exoskeleton augments are shoulder–elbow,
shoulder–elbow–wrist, or shoulder–elbow–wrist–hand.
d. DoFs: the degrees of freedom of the analyzed exoskeletons vary from 1 to 7, and they
can be active or passive depending on the type of mechanism used.
e. Rehabilitation type: in terms of rehabilitation modalities, there are several approaches,
such as active and/or passive assistance provided by the exoskeleton.
It has also been observed that there is very little evidence of clinical testing in the
cardiac rehabilitation of existing exoskeletons, raising doubts about the already developed
exoskeletons’ true efficiency and utility. In this context, in which there is an acute need for
cardiac and COVID-19 rehabilitation, and, according to the World Health Organization, the
fact that the number of people aged 65 and over will rise by 73% in developed countries
and 207 percent globally by 2050 [30], it is necessary to find new solutions to innovatively
solve the acute shortage of rehabilitation specialists and increasing addressability of CR.
This situation must be resolved, especially since this age group is at a higher risk of cardio-
vascular disease, and CR can help them improve their quality of life. This solution must
come from the direction of the development of robotic exoskeletons for CR rehabilitation.
Following cardiac surgery, patients experience difficulties reaching an acceptable level
of independence, even for simple physical activities [31]. The rehabilitation process is
long and painful, and often lacks challenges and fun, especially in an unfavorable context
(e.g., depression or difficulty attending a rehab centre) [32]. Many patients may thus lack
the motivation for rehabilitation activities. Even in developed countries, the number of
cardiac patients participating in rehabilitation activities is very low (e.g., 13 percent in
Denmark) [32].
Furthermore, the scientific literature increasingly reports the contribution of extrinsic
motivation mechanisms (serious games/exergames) to rehabilitation effectiveness [33].
Extrinsic motivation, via gamification, brings fun and joy to rehabilitation activities and
diverts attention from inherent pain [31].
Mobile networks and the internet can make up for the distance between patients and
physiotherapists; thus, secondary rehabilitation stages can be performed by the patient
at home, with remote monitoring. Motivation makes people do particular actions, giving
specific reasons for these actions or needs. Thus, including motivation mechanisms in
rehabilitation, both in the hospital/clinic and at home, can boost the number of patients
performing complete rehabilitation.
Gamification (serious games) means inserting game mechanics into the non-game
environment [34]. It can increase user engagement, as people enjoy an interactive process
full of fun, challenges, and competitive spirit, similar to games. Curiosity and excitement
drive people to continue performing various tasks and to spend more time on the system
they interact with. Technically, gamification leads to improving the KPIs of a process
(rehabilitation, in our case) [35]. These KPIs are important in defining the game rules and,
more importantly, in communicating to patients what the rehab goals are.
In this paper, we present the design of a full upper limb robotic exoskeleton augmented
by a virtual reality (VR) non-immersive module for the rehabilitation of patients following
open-heart surgery or a major cardiac event. The proposed system provides rich sensory
feedback that is similar to that of a natural and typical real-world experience, and it has
been designed for scenarios including goal-directed activities such as games (e.g., virtual
trainer, and picking up and placing different objects). To encourage the patient to conduct
self-initiated and regulated motions, a virtual trainer assistance was introduced.
Symmetry 2022, 14, 81 5 of 30

2. Materials and Methods

2.1. CardioVR-ReTone Design Overview
The CardioVR-ReTone robotic exoskeleton, designed in this paper, is presented in
Figure 1. The CardioVR-ReTone symmetric exoskeleton offers a total of twelve DoFs: six
DoFs on each arm. The shoulder joint is represented by the first five motors, whereas the
elbow joint is represented by the sixth motor. Being symmetric, the CardioVR-ReTone
exoskeleton facilitates symmetric movements, which will have a beneficial effect on the
functional upper extremities and reduce the injury risk.

Figure 1. The designed CardioVR-ReTone exoskeleton. Isometric 3D view (left); kinematic struc-
ture (right).

The shoulder mechanism was designed as a rotational joint. The J1 rotation joint
(Figure 1) mimics shoulder raising and lowering mobility, while the triangle J2–J3–J4
mimics shoulder protraction and retraction mobility. The kinematic discrepancy that would
be caused if only one rotation joint was employed for protraction and retraction on the back
side of the shoulder is eliminated by arranging the joints in a triangle.
Two revolute joints positioned at 90◦ were used to create the forearm mechanism. This
technique was chosen for the forearm mechanism in order to improve ROM while avoiding
mechanical singularities and upper arm movement obstacles. Axis J5 is oriented to the axis
X, while joint J4, which is shared with the shoulder mechanism, is aligned to the vertical
Cartesian axis (Figure 1).
The elbow mechanism consists of two revolute joints, one of which is active and
actuated by a motor (Figure 1—joint J6) to allow for elbow flexion and extension.
The CardioVR-ReTone exoskeleton’s two robotic arms are attached on a height- and
width-adjustable frame, allowing it to fit a wide range of patients’ sizes (Figure 1).
The resultant CardioVR-ReTone is a simple, integrated, and morphologically compati-
ble exoskeleton with mass and volume dispersed over the robotic exoskeleton framework.
Abduction and adduction, flexion and extension, horizontal abduction and adduction, and
internal and exterior rotation are some of the movements that the exoskeleton can assist
and perform with the human upper torso and arms.
The movements of the CardioVR-ReTone system were designed to comply with a
medical early CR protocol. The European Society of Cardiology (ESC) Guideline for CR [36]
recommends that upper-body training can begin when the chest is stable, i.e., usually after 6
weeks. In our protocol, we proposed to start the intervention earlier during hospitalization
and to compare the results obtained after usual CR with those obtained using early CR
assisted by a robotic device. We chose this type of protocol because there are studies that
showed that early training after cardiac surgery is beneficial for the patient. We propose
Symmetry 2022, 14, 81 6 of 30

early CR based on the meta-analysis by [37], which comprised the best studies available
in the literature, and gave provisional substance to the general belief that aerobic exercise
commenced early after cardiac surgery (during the first postoperative week) significantly
improved functional and aerobic capacity, with a rate of adverse effects that was relatively
low and not different to usual CR. The protocol of standardized exercises was refined by the
medical team in collaboration with experienced kinetotherapists in accordance with the lat-
est rehabilitation guidelines [36] and international recommendations [38]. The standardized
protocol is available in the supplementary files of our group’s recent publication [39].

2.2. Forward Kinematics

The geometric and kinematic model for the CardioVR-ReTone exoskeleton was devel-
oped for the right arm. Further, using dedicated algorithms, the direct and inverse geometry
and kinematics equations were determined for the CardioVR-ReTone exoskeleton—right
arm 6R mechanical structure. The specific transfer matrix equation for a mechanical struc-
ture having rotational (R) or translational (T) motions, according to [40], can be expressed
using algorithms such as a matrix of locating algorithm. The obtained results are essential
to the optimal design, dimensional aspect, and energy, but also to simulate the kinematic
and dynamic behavior of the CardioVR-ReTone exoskeleton. This paper considers the
terms “kinetic link” and “joint” to have the same meaning.

2.2.1. Geometrical Modeling of the CardioVR-ReTone Exoskeleton

The direct geometry equations (DGM equations) can be determined by applying the
matrix of locating algorithm, considering the minimum number of geometric or mechanical
restrictions. According to [41], the situation matrices are defined as:
" # " #
(0) (0)
R i i −1 i −1 p i i −1 Ri 0 pi 0
Ti i−1 = ; Ti 0 = (1)
000 1 000 1

Similarly, for i = n + 1, the locating matrices between the frames {n} → {n + 1} and
{0} → {n + 1} are defined according to the following:
" #
(0)
n (0) s (0) a (0) n p n +1 n
Tn+1 n =
0 0 0 1
 (0) (0) (2)
a (0) p (0)

n s
Tn+1 0 = Tn 0 · Tn+1 n =
0 0 0 1

In expression (2), p(0) is a vector that defines the position of the last joint with a
(0)
reference system attached to base of the robot, and n p n+1 n characterizes the relative position
of the system attached to the end effector beside the geometrical center of the last joint.
For i = 1 → n , the situating matrices between the two close related systems {i − 1} → {i }
are defined as: "   #
  R ki ; qi · ∆ i (1 − ∆ i ) · q i · i k i
T∆ ki ; qi = ; (3)
0 0 0 1
 
i −1
i [ T ] = Ti i − 1 · T∆ k i ; q i . (4)

According to [41,42], the rotation matrix between two neighboring reference frames is:
i −1
i [ R] = { R ( x i ; qi · ∆ i ) ; R ( yi ; qi · ∆ i ) ; R ( zi ; qi · ∆ i ) } (5)

i −1 (0)
r i = i −1 p i i −1 + (1 − ∆ i ) · q i · i k i . (6)
Symmetry 2022, 14, 81 7 of 30

For i = 1 → n , the position vector between {i } and {i − 1} with respect to {0} fixed
frame, the position of joint {i } related to the same fixed frame is determined as:

i
pi i−1 = i−01 [ R] · i−1 r i i−1 ; pi = ∑ p j j −1 (7)
j =1

The situation matrix between {0} → {i } is:

i  0 [ R]

j −1 pi
0
i [T ] =∏ j [T ] = i . (8)
j =1
000 1

and for the end effector:

 
0 0 n s a p
n +1 [ T ] = n [T ] · Tn+1 n = . (9)
0 0 0 1

The orienting vector is defined [41,43] with the following identity:

n o
R(α A − β B − γC ) = n+01 [ R] ; ψ = (α A β B γC ) T (10)

The DGM equations are included in the following generalized matrix:

   T 
p p x py pz
0
X =  − − −  =  − − − − − − − . (11)
 
 T
ψ α x β y γz

The direct geometrical modeling, regardless of the algorithm used, aims to establish
the geometry equations that will serve in determining the direct kinematic model (DKM).
(0)
According to the input data corresponding to DGM, the matrix of nominal geometry Mvn
for the 6R CardioVR-ReTone robot is presented in Table 2.

Table 2. Nominal geometry for the 6R CardioVR-ReTone robot.

(0)
Mvn ∈6R (Right)
Joint pi ki
Joint Type
i=1→6 {R ; T} @pxi @pyi @pzi kxi kyi kzi
1 R 0 l0 l1 0 1 0
2 R 0 − l2 l3 0 0 1
3 R l4 − l5 0 0 0 1
4 R 0 l6 0 0 0 1
5 R l7 0 − l8 1 0 0
6 R − l9 0 −l10 1 0 0
7 - 0 l11 0 1 0 0

Figure 2 highlights the kinematic scheme of the CardioVR-ReTone robotic exoskeleton.

In keeping with Table 1 and Figure 2, Table 3 highlights the geometrical particularities for
the 6R CardioVR-ReTone robot.
Symmetry 2022, 14, 81 8 of 30

Figure 2. The kinematic scheme of the CardioVR-ReTone robotic exoskeleton 2x(6R).

Table 3. Geometrical particularities for the 6R CardioVR-ReTone robot.

Joint Right Left Kinetic Link Orientation Vector

i ∆i = 1 k i ≡ { x i , yi , zi }
 T
1 y ki ≡ yi = 0 1 0
 T
2, 3, 4 z ki ≡ zi = 0 0 1
 T
5, 6 x ki ≡ xi = 1 0 0

For i = 7, the situation matrices between frames {7} → {6} are:

 
0 1 0 0
6
 0 0 1 l11 
7 [T ] ≡ T76 =
 1 0
; (12)
0 0 
0 0 0 1

which, in keeping with (8), according to (9), lead to:

 
0 [ R]
 
0 n s a p7 p7 
= 06 [ T ] · 67 [ T ] =

7
7 [T ] = ; (13)
0 0 0 1  
0 0 0 1

representing the resulting orientation matrix and the position vector, between {7} → {0}
frames, both being included in the expression of the column vector of operational variables (11).
Symmetry 2022, 14, 81 9 of 30

According to [41], in order to establish the orientation angles α x , β y , γz for exact

determination of the values, the trigonometric function A tan 2 is used, defined by:
 
{α; [sα ≥ 0; cα > 0]};

 

{π/2 + α; [sα > 0; cα < 0]}
 
x = A tan 2(sα; cα) = (14)

 {π + α; [sα < 0; cα < 0]}; 

{−π/2 + α; [sα < 0; cα ≥ 0]}
 

Hence, in keeping with (14), results in:

nπ πo nπ πo
αx = ,− , β y = {π, 0}, γz = ,−
2 2 2 2
The column vector of operational coordinates, defined by (11), becomes:
h . . iT
0
X= p6 π π π (15)
2 2

The expression (15) characterizes the direct geometric modeling of the robot type 6R
studied. The parameters included in (15) express the position and orientation of the end
effector with respect to the fixed reference system, attached to the base of the robot.

2.2.2. Development of Kinematical Modeling for CardioVR-ReTone Exoskeleton

The operational kinematic parameters that express the movement of the end effector
in Cartesian space will be discussed in this paragraph. To define these parameters, the
iterative algorithm is used [41]. This algorithm defines the kinematic parameters with
respect to the fixed reference frame, attached to the base. The kinematic analysis considers
the position and orientation of each link necessary to describe the location of the end
effector in the robot workspace [42,44].
In applying the iterative algorithm [41], in a first step, for the first joint J1 (i = 1), it
must be assumed that the absolute kinematic parameter values corresponding to the fixed
base of the mechanical structure of the robot are:
n . . o
0
ω 0 = 0, 0 ω 0 = 0, 0 v 0 = 0, 0 v 0 = 0 .

For i = 1 → n , the angular and linear velocities that define the absolute motion of
each kinetic link are determined using the expressions:
.
0
ω i = 0 ω i−1 + ∆ i · 0i [ R] · qi · i k i (16)
.
 
0 0
vi = v i−1 + 0 ω i−1 × pii−1 + (1 − ∆ i ) · qi · 0 k i ; (17)

Similarly, for each i = 1 → n , the kinetic link of the robot, the corresponding angular
and linear accelerations projected on the fixed reference system are defined as follows:
. . n . ..
o
0
ω i = 0 ω i−1 + ∆ i · 0 ω i−1 × qi · 0i [ R] · i k i + qi · 0i [ R] · i k i (18)

.  . . 
0v = 0v + 0 ω i−1 × pii−1 + 0 ω i−1 × 0 ω i−1 × pii−1 +
i i −1
n . ..
o (19)
+(1 − ∆ i ) 2 · 0 ω i × qi · 0 k i + qi · 0 k i

In the second part of the iterative algorithm, the kinematic parameters, which char-
acterize the movement in each i = 1 → n kinetic link relative to the {i } mobile reference
system, are determined as:
.
i
ω i = 0i [ R] T · 0 ω i = i−1i [ R] · i−1 ω i−1 + ∆ i · qi · i k i ; (20)
Symmetry 2022, 14, 81 10 of 30

iv = 0i [ R] T · 0 v i = i−1i [ R] · i−1 v i−1 + ·i−1 ω i−1 × i−1 p ii−1 +


i
. (21)
+(1 − ∆ i ) · qi · i k i
The expressions of angular and linear accelerations projected on the mobile reference
system {i }, that characterize the relative movement of each i = 1 → n link, are defined by
the following expressions:
. . .
iω 0 0 i i −1 ω
i = i [ R ] · ω i = i −1 [ R ] · i −1 + o
n
− . .. i (22)
+ ∆ i · i −1 [ R ] · ω i −1 × q i · k i + q i · k i
i i 1 i

. . n . .
iv
i = i−1i [ R] · 0 v i = i−1i [ R] · i−1 v i−1 + i−1 ω i−1 × i−1 p ii−1 +
+ i−1 ω i−1 × i−1 ω i−1 × i−1 p ii−1 +  (23)
. ..
+(1 − ∆ i ) · 2 · i ω i × qi · i k i + qi · i k i

In the last step of the iterative algorithm, the absolute motion of the final effector i = n
is defined, considering the situation equations, respectively the absolute linear and angular
velocities and accelerations (operating velocities and accelerations).
. h iT
( n )0 ( n )0 v T ( n )0 ω T
X= n n ; (24)

.. h . . iT
( n )0 ( n )0 v T ( n )0 ω T
X= n n . (25)

Taking into account the relations (24) and (25), the following is obtained:
0v
 
0
.
. 6
X ≡  − − −  = 0J θ · θ (26)
0ω
6
 . 
0v
.. 6 ..
0
X ≡  − − −  = 0J θ · θ


(27)
 
.
0ω
6

The expressions previously determined are the direct kinematics equations (DKM)
that characterize the absolute motion of the end effector, and will be used as input data in
the dynamic modeling [43] and representation of the equations of direct kinematics, that
characterize the motion of the final effector (speed and acceleration) of the 6R robot in the
Cartesian space, where in (26) and (27), 0 J θ is the Jacobian matrix, as:

0 0J 0J 0J 0J 0J 0J

 
J θ = 1 2 3 4 5 6 (28)
(6×1)

Based on the same algorithm, the kinematic control functions of the serial structure of
type 6R are determined, as [44]:
.  −1 0 .
θ (t) = 0 J θ (t)

· X (t) (29)
.. −1 0 ..  −1 0 .   .
θ (t) = 0 J θ (t) · X (t) − 0 J θ (t)
 
· J θ (t) · θ (30)
 −1
where 0 J θ (t)

is the inverse of the Jacobian matrix.
Taking into account Equations (24) and (25), the operational speeds and accelerations
can be expressed compared to the own system as:
.  6v

6 6
X= 6ω (31)
6
Symmetry 2022, 14, 81 11 of 30

" . #
.. 6v
6 .6
X= 6ω
(32)
6

and define the motion of the joint with respect to its own reference frame.

2.3. Actuators, Electrical, and Electronic Design

Exoskeleton movement is a delicate control procedure, especially in the subfield of
human rehabilitation after open-heart surgery. Actuators are the prime movers in any
mechanical system, and they are capable of transforming different types of energy into
movement or force. This section discusses the identification of the actuators and sensorial
systems used to generate motion in biomechanics and sensorial systems. Considering the
developed kinematic model and mechanical design, we will calculate the requirements of
the actuators and trigger the selection mechanism.
Several actuator types are used in upper limb exoskeletons: electric motors coupled
with gearboxes; cables driven by electric motors; hydraulic cylinders; pneumatic muscles
actuators (PMA); shape memory alloys (SMA); and series elastic actuators (SEA).
The development of a control system suitable for this type of application involves an
analysis of actuator and sensorial systems used previously by fellow researchers.
As identified in the scientific literature, electric motors in combination with a gearbox,
preferably a harmonic drive gearbox, are used the most often in exoskeletons. Feedback
systems can be divided in two categories: human feedback systems and exoskeleton
feedback systems.
Human feedback systems are used to predict human desired motions and use sensorial
systems like surface electromyogram (sEMG), electroencephalogram (EEG), electrooculo-
gram (EOG), electrocardiogram (ECG), muscle circumference sensors, inertial measurement
unit (IMU), force sensors, and gesture detection by means of video analysis. sEMG are the
most commonly used type of human sensorial system.
Exoskeleton feedback systems are used to ensure the correct control loop of the
actuators and use sensorial systems like encoders for detecting the direction, speed, and
position of a joint, torque control, potentiometers, joint angles, flexion sensors, pressure
sensors, and vision tracking and VR systems. Most often, the exoskeleton feedback loop
considers torque control and encoder feedback. VR systems have attracted attention due to
their capacity to motivate human subjects.
After considering all of the actuator types, the BLDC electric motor coupled with
harmonic drives was chosen due to its good torque-to-weight ratio, compact size, low
backlash, and ease of control. Even though pneumatic muscles actuators (PMA) and
shape memory alloy (SMA) have the great properties of reduced weight, there are major
constraints related to the control techniques and team experience with such actuators
compared to the control and integration of electric actuators. The characteristics of the
harmonic drive complete electric actuator series, consisting of an EC60 flat BLDC Maxon
motor, optical encoder, and HD gear, are presented in Table 4.

Table 4. Harmonic drive complete actuator’s characteristics.

Characteristic CPU-14 CPU-17 CPU-20 CPU-25 CPU32

Gear ratio 100 100 100 100 100
Repeated peak torque (Nm) 28 54 82 157 333
Average torque (Nm) 11 39 49 108 21
Rated torque (Nm) 7.8 24 40 67 137
Momentary peak torque (Nm) 54 110 147 284 647
Gear weight (kg) 0.54 0.79 1.3 1.95 3.9
Motor weight (kg) 0.5 0.5 0.5 0.5 0.5
Symmetry 2022, 14, 81 12 of 30

Torque Requirements Calculation

Starting from the kinematic model highlighted in this article, and taking into consid-
eration the information regarding the properties of the exoskeleton segments, as well as
the anthropomorphic characteristics of the upper limb body segments for Romanian male
subjects and actuator features, we then moved on to the selection of electric motors. Table 5
presents an overview of the mechanical properties of exoskeleton links as obtained from the
design phase using the Catia© 3D modeling software. Every link connects two exoskeleton
joints, as mentioned in the subscript. Joints J2 and J3 are interconnected by link L2,3 . Table 6
presents an overview of the upper limb anthropomorphic characteristics of Romanian male
subjects, and defines a relation between the exoskeleton joints that have to sustain the load
introduced by a specific body segment.

Table 5. Exoskeleton segments.

Link between Joints Length Center of Gravity Weight

Joint Number j
Lj,j+1 (cm) (cm) (kg)
1 L1,2 15 7.5 0.3
2 L2,3 19.8 8.11 0.427
3 L3,4 35.63 15.61 0.847
4 L4,5 17.3 8.64 0.292
5 L5,6 37.2 17.57 1.209
6 L6,E 22.170 15.87 0.923

Table 6. Mean values of anthropomorphic body segments.

Center of
Exoskeleton Body Segment Length Weight
Body Segment Gravity
Joint Groups EJG Number j (cm) (kg)
(cm)
1: J1- > J2 1 Shoulder 14 7 3.47
2: J3- > J5 2 Upper arm 29.425 14.92 3.73
3 Forearm 28.375 11.83 2.06
3: J6
4 Hand + joystick 10.125 5.17 1.0

The actuator selection considers calculating the torque required by a specific joint
to maintain equilibrium with the rest of kinematic chain, including previously selected
actuators and associated human body segments. It was calculated using the following
formula (33):
Tj = ( Tjexo + Tjbody ) × Ks (33)
where j is the joint for which torque is calculated, Tjexo (34) represents the torque required by
a specific exoskeleton joint to maintain equilibrium with the associated exoskeleton segment,
Tjbody (34) represents the torque required by a specific exoskeleton joint to maintain equilib-
rium with the associated human body segments, and Ks is an oversize safety coefficient.
  
joints joints−1 joints
Tjexo = ∑ j LinkLength j × Link CGj × FGj + ∑ j ∑k= j+1 LinkLength j ×

joints−1

joints−1

joints
 (34)
FGk + ∑ j ∅ Mj+1 × CGMj+1 × FGMj+1 + ∑ j ∑k= j+1 LinkLength j × FGMk

where,
LinkLength j = length of the link between exoskeleton j and j + 1
Link CGj = distance from the starting point of the link up to link j center of gravity
FGj = gravity force acting on a link j
Symmetry 2022, 14, 81 13 of 30

∅ Mj = diameter of actuator that drives joint j

CG Mj = center of gravity of the actuator driving joint j
FGMj = gravity force acting on actuator j
  
segments−1 segments
Tjbody = ∑ j= EJ ∑ k = j + 1 SegmentLength j × FGk +
G (35)
segments
∑ j= EJG SegmentLength j × Link CGj × FGj

SegmentLength j = length of the human body segment j

EJG = represents the joint group number associated with the movement of the
body segment.
The actuator selection started from joint J6 to joint J1. This approach is required as the
actuators selected for a joint will involve additional torque requirements for the actuator
selection process of previous joints. Table 7 presents an overview of the torque and safety
coefficient estimation, together with the actuator selection for every exoskeleton joint.

Table 7. Torque estimation and actuator selection.

Actuator
Tjexo Tjbody Tj
Joint Ks Average Repeated Peak Weight
(Nm) (Nm) (Nm) Type
Torque (Nm) Torque (Nm) (Kg)
6 1.44 5.69 1.25 8.91 CPU-17 39 54 1.29
5 12.09 19.97 1.25 40.07 CPU-17 39 54 1.29
4 20.95 19.97 1.25 51.15 CPU-20 49 82 1.8
3 46.90 19.97 1.5 100.30 CPU-25 108 157 2.45
2 64.76 31.68 1.5 144.66 CPU-25 108 157 2.45
1 88.96 31.68 1.25 150.80 CPU-25 108 157 2.45

As can be observed in Table 7, with the exception of joints J1 and J2, all joint torque
requirements are smaller than the average torque produced by the selected actuator. This
means that the actuator has enough torque to control the exoskeleton and the attached body
segments. Joint 4 is close to the average torque, but low under repeated peak torque value.
The torque requirement for joints J1 and J2 is greater than the average torque, but still
under the repeated peak torque value. Even though a CPU-32 actuator would have been a
more reliable option, its dimensions and weight together with financial factors constrained
the selection process for CPU-25.
After the actuators were selected, identifying the right actuator motor controller was
required in order to meet the technical and application requirements. For this application,
every actuator motor was controlled by a Maxon EPOS4 5/50 controller with a CANopen
fieldbus interface. The EPOS4 5/50 controller can provide speed control, position control,
and torque control loops. On top of these control features, the EPOS4 controller can be
easily integrated within high programming languages or other PC-based software with the
help of several plug-ins provided by the manufacturer.

2.4. Control Architecture Design

The control architecture for the CardioVR-ReTone exoskeleton, presented in Figure 3,
consists of several important components: virtual reality, a Simulink environment, a
main control unit, an exoskeleton actuation system, an exoskeleton control joystick, and
sensorics. The virtual reality component consists of an application that will immerse the
subject within a virtual environment to enhance the rehabilitation experience by triggering
motivational mechanisms.
Symmetry 2022, 14, 81 14 of 30

Figure 3. CardioVR-ReTone exoskeleton control architecture.

Simulink is the environment where kinematic equations are solved and the exoskeleton
joint coordinates, speed, and torque are calculated, taking into consideration the selected
exoskeleton operation and commands triggered by means of an exoskeleton joystick.
The main control unit is a fail-safe open controller developed by Siemens that combines
the functionalities of programmable logic controllers with a PC-based platform preinstalled
with Windows 10 IoT Enterprise. The Simatic ODK 1500 software controller that runs on
the hardware of the open controller can provide a seamless integration and connectivity
of high-end programming languages with open controller hardware resources. Therefore,
the virtual reality, robot operating systems (ROS), and Simulink software are installed and
executed on the open controller. The data received from Simulink are sent to the open
controller hardware and delivered to every exoskeleton joint actuator controller using a
CANopen communication module.
The exoskeleton actuation system is built out of several subcomponents: the HD gear,
the motor, the Hall sensors and encoder, and the digital position controller EPOS4. The
EPOS4 controller drives the exoskeleton joint, considering the motion parameters calculated
by Simulink and delivered by the main control unit via the communication protocol to
each EPOS4 controller. The EPOS4 controllers are enhanced with a safe torque off (SFO)
functionality, which is triggered by pressing the emergency button on the joystick. The SFO
turns off the power on the controller output to prevent the motor from producing torque.
The EPOS controller uses the optical encoder and Hall sensors within the positioning
and speed control loops. The torque control loop considers the current injected into the
motor’s windings.
Sensorics components are the surface EMG electrodes placed on the human body
that can identify the subject’s muscle activity. The muscle activity values are fed into the
Simulink algorithm using Bluetooth, calculating how much torque the actuators need to
compensate for to reach the desired position within the rehabilitation exercise.
Symmetry 2022, 14, 81 15 of 30

An exoskeleton control joystick is used for operating mode selections. Three operating
modes are considered for development: assisted, partially assisted, and resistive. Assisted
mode is the operating mode in which the exoskeleton actuating systems move the subject’s
arms using the exoskeleton structure without any major effort for the operator, according
to the motion parameters generated by Simulink with respect to the rehabilitation exercise
selected. The partially assisted mode is an operating mode whereby the EMG sensors’
feedback is considered by the Simulink algorithm, and only a specific amount of torque
is generated by the motors to compensate for the subject’s actions towards reaching the
desired position. The resistive mode is the operating mode in which the exoskeleton
actuating system opposes the subject’s actions in order increase the subject’s effort towards
achieving a specific action. In every operating mode, the patient can use the joystick
buttons to increase or decrease the operating speed and torque delivered by the exoskeleton
actuating system during their rehabilitation exercises. In addition, by means of the joystick
buttons, the subject or the medical staff can select a specific exoskeleton joint and move it
forwards or backwards in order to customize the subject’s rehabilitation exercises and fit
the exercises to the user’s actual body condition.
In terms of safety features, the exoskeleton actuation system can be powered only
if the emergency stop buttons are not pressed. Emergency stop buttons are mounted on
every exoskeleton arm, and a third emergency button is mounted on the base frame of the
exoskeleton. The emergency stop buttons will trigger the SFO functionality. Another safety
feature that turns the actuators on/off is the dead-man switch, not represented in Figure 3.
The exoskeleton system only follows the motion controls provided by Simulink when the
dead-man switch is placed in an intermediary position. If the dead-man switch is released
or pressed too hard, the actuators stop.
The exoskeleton control loop (see Figure 4) is based on a three-level cascade PI(D) loop,
which is a common practice in most of the control systems involving motion control. The
main control loop receives the position, velocity, and acceleration from the path planner,
controls the position of the joint actuator, and provides motor velocity and acceleration as
outputs. The inner loop controls the actuator velocity and outputs the current reference for
the innermost control loop. The encoder mounter on the BLDC motor feeds information
for the speed and position control loops.

Figure 4. Exoskeleton control loop.

Within the current control loop or torque control loop, since torque is directly propor-
tional to the current through the motor windings multiplied by the motor constant, partial
torque is delivered to achieve the desired position with a desired speed. In the partially
assisted operating mode of the exoskeleton, the control loop is completed with a torque
estimation of human muscle based on a body model which is fed with EMG signals.
Symmetry 2022, 14, 81 16 of 30

The VR motivational application requires process information from the main control
unit to detect if the subject is following the rehabilitation exercise pattern, and to identify any
error between the actual versus the desired arm position in order to adjust the motivational
exercise intensity. Such process information can be related to exoskeleton TCP which
includes coordinates on X, Y and Z and orientation angles. Calculating exoskeleton TCP is
based on feeding information gathered from the motor encoders to the direct kinematic
model that runs on Simulink. An ROS node, that makes use of the MODBUS TCP fieldbus
(modbus_plc_siemens), cyclically reads exoskeleton TCP from the main control unit and
feeds the data to an ROS node that manages subject movements to detect whether the arm
is moved according to the rehabilitation exercise.

2.5. Patient Extrinsic Motivation with VR

2.5.1. Operating and Control Application: Interaction Scenarios
To design the interaction between the patient and the exoskeleton-based rehabilitation
system, we built interaction scenarios that cover the entire rehabilitation process lifecycle
(starting immediately after the surgery, continuing at the patient’s home, and ending once
the patient is completely recovered). The first rehabilitation stage covered by the interaction
scenarios begins with the first movements the patient has to make, after the surgery. The
patient will be (fully) supported by the exoskeleton in performing these movements, based
on physical exercise templates established by the therapist. In the second rehab stage, the
patient performs the movements being assisted by the exoskeleton (e.g., they will need to
put in some effort, just being helped by the exoskeleton). The CardioVR system will check
the movement “correctness” (how “far” the movements are, compared to the movement
template set by the therapist). The third stage is after discharge, when the patient performs
exercises without any support and only being monitored by sensors (Android wrist bands
and/or a Kinect sensor).
The motivation (gamification) mechanism we used in the CardioVR software plat-
form is the patient’s rehabilitation journey, comprising a series of levels to be carried out
(steps of recovering mobility) up to complete rehabilitation. These levels are decided by
a rehab mentor (e.g., the therapist), who will set up the patient’s journey and will assist
them throughout.
The patient will follow the rehab journey, which will be managed by the software
platform. The patient will be shown the steps to take, their current physical state, the
short-term mobility goals, current level challenges (why the patient has to do it and why
it may be difficult), and also interactive lessons on cardiac disease, post-surgery rehab,
healthy lifestyle (e.g., “what to know to avoid future health issues”). This will be done
via a VR game (immersed or not, depending on the patient’s age and potential aversion
to “unknown” technology). The patient will receive the challenges in stages, according to
their rehab progress. A level consists of a series of physical exercise sets to be performed.
For each level, the “software” mentor in the VR game will explain the exercises and the
challenge. Regardless of the rehab level, there are three user roles (patient, therapist,
therapist assistant) and two interaction scenarios.
Scenario 1: Focused on the therapist, it allows a healthy user (a therapist assistant) to
perform a set of rehab exercises, using the exoskeleton, for “recording” purposes, to create
an exercise template. The software platform records these movements (e.g., it records all
the movement parameters (angles and velocities) against a time reference and then stores
them). In this way, a template of “correct” physical exercises can be created. This can be
general or tailored to a specific patient need. The template is then attached to the patient’s
rehab journey, as a level.
Scenario 2: Focused on the patient, it allows the selection of their current rehabilita-
tion task (level) and getting the basic information (“why do I have to do this?”) and the
motivation (e.g., bonus points to gain). The patient then performs the exercises, either
assisted or supported by the system, which monitors how correctly the movements are
Symmetry 2022, 14, 81 17 of 30

made. The correctness is determined by comparing the movement parameters against the
stored template.

2.5.2. Operating Software Architecture

The architectural diagram of the CardioVR-ReTone operating platform is presented in
Figure 5, and was constructed using a distributed control approach (to be implemented
over the ROS framework [45]). The two user symbols depict the recovering patient (see
Scenario 2 above) and the physiotherapist (see Scenario 1 above).

Figure 5. CardioVR-ReTone operating and control platform logical modules.

The exoskeleton sensors and actuators are controlled at a logical level by the “sensors”
and “actuators” modules, which act as hardware drivers. These modules communicate
with the “play” and “rec” modules via asynchronous messages. The (asynchronous)
communication mechanism is coordinated by a message handler.
The “rec” module records the parameters of the sample rehabilitation exercises ex-
ecuted by the therapist assistant (see Scenario 1 above) and stores them in the DB as a
“sample”. The “play” module moves the exoskeleton according to a sample of exercises
and thus supports the patient in the required arm movements. The “set” module allows
the physiotherapist to build the rehab journey of a patient (e.g., what exercises to perform
in each rehab level) and store it in the database.
In the second interaction scenario, as the patient executes the rehabilitation exercises
within a rehabilitation level, the communication node of the “play” module requests the
movement parameters (X, Y, Z, and orientation angles of the TCP) from the exoskeleton
control unit, via messages over a dedicated communication topic, with a 10 Hz frequency.
Symmetry 2022, 14, 81 18 of 30

Based on this data, the “play” module counts the movements done by the patient (by
comparing them against the reference, i.e., the stored exercise template). It further sends the
counter data to the VR game and to the “exoskeleton-as-keyboard” module. The VR game
uses this data to visually correlate the avatar movements with the actual patient movements.
If an external game is used as a motivator within a rehab level, the “exoskeleton-as-
keyboard” module translates this data to virtual key presses that, in turn, control the
external game (e.g., executing jumps, left or right turns).
The “settings module” is the business logic corresponding to the settings user in-
terface, and the “VR module” is the business logic of the VR game. In the partially and
resistive assisted scenario, an external motivational game (e.g., any endless runner or
moto-race simple open-source game) can be played by the patient; to control the game, the
patient movements are translated into virtual key presses by the “exoskeleton-as-keyboard”
adapter. For instance, to control an endless runner game, one arm raise can be translated to
a “jump” (key up).

2.5.3. Gaming Components

Two “game” components implement the motivational mechanisms discussed above:
the VR game that accompanies the patient along their rehab journey (like the “main menu”
of a game where you see/select the current level to play), and other (simple) games to
accompany the patient along a single level. These are discussed below.
The VR game: to make the entire rehab journey easily understandable and fun, a VR
game is being developed to act as an avatar “host” for the patient. It conveys information
such as the current progress, what is coming next, and medical information related to the
patient’s disease, cure, and recommended lifestyle. When presenting the next exercise level,
the avatar explains the movements to be made and trains the patient to do them, while also
providing feedback.
The hardware used to track the patient’s hand movements has been developed to
make use of the Microsoft Kinect V2 motion sensing device.
Many researchers [46] have started to develop various Exergame VR applications
that make use of widely available and affordable motion sensors such as Microsoft Kinect,
which have started to be used in various physical recovery activities [47]. This sensor offers
an accurate tracking system that can provide reliable posture analysis regarding the angular
constraints of the patient’s hand movement. Thus, the patient will be encouraged by the
VR exergame to define the required hand recovery movement.
The validity and reliability studies that make use of Microsoft Kinect V2 are widely
available in the literature, where researchers have compared the real-time skeleton tracking
system with advanced 3D motion tracking systems such as Vicon or as conventional clinical
goniometers and digital goniometers. Other researchers [30,48] have compared the body
movement recordings of both Vicon and Kinect datasets to identify movement patterns
during exergaming. Other researchers have highlighted the current implications of virtual
reality applications intended for cardiac rehabilitation exergaming [30]. These virtual reality
applications have a promising impact to improve patient self-efficacy for exercise training
using various digital hardware setups. The mostly common setups make use of affordable
and widely available equipment such as a Kinect or various virtual reality head-mounted
display (HMD) headsets. Various arm movements such as abduction, adduction, flexion,
extension, internal, and external rotations have been analyzed. These are illustrated in
Figure 6.
Symmetry 2022, 14, 81 19 of 30

Figure 6. The movements of the CardioVR-ReTone exoskeleton necessary for CR.

The level games: to build enthusiasm, introduce fun into a rehab level, and to distract
from potential pain, a simple game can be played while performing the exercises. To avoid
boredom and to be able to adapt to each patient, any game (for which a valid license exists)
can be used in this step.
The patient’s movements, either free or exoskeleton-assisted, are translated into key-
presses: for instance, raising the arms can be equivalent to the “up” key or the difference in
movement between the arms can correspond to the “left” or “right” key. Simple voice com-
mands like “now” or “fire” can also be used as the “enter” or “space” key. At runtime, the
“exoskeleton-as-keyboard” module sends the “translated” key presses to a computer run-
ning the external game. In the current software version, the Linux xdotool [34] command
line tool was used to simulate keyboard activity.
The movements described in Figure 6 are in accordance with the CR protocol de-
scribed in our previous work [39] and in accordance with the available CR guidelines and
international recommendations [36].
Symmetry 2022, 14, 81 20 of 30

3. Results
The dynamic control functions will be established for two kinetic joints—J4 and J5
(Figure 1). Based on the algorithm of the generalized forces in dynamics, the general-
ized driving forces will be determined using analytical calculations and are graphically
represented in the sections below.

3.1. Differential Motion Equations for CardioVR-ReTone—6R Structure

According to [41,42], differential equations of motion can be determined based on
Euler–Lagrange equations, since the Euler–Lagrange formalism is specific to non-conservative
mechanical systems with holonomous links. In the paper, the equations of motion for two
driving joints, J4 and J5, are determined for the CardioVR-ReTone 6R robot. The two joints
are highlighted in Figure 7.

Figure 7. The CardioVR-ReTone mechanical structure and joints range.

In order to establish the motion expressions for the considered links, the driving
moments will be established as:

Qim = QiiF + Qig + QiSU , i = 4, 5 (36)

n o
Previous expressions include QiiF ; Qig ; QiSU , known as: generalized inertia forces,
generalized gravitational forces, and generalized handling forces [43].
The expression for generalized inertia forces for i = 4, 5 is determined as:
  .  .
∂ E θ; θ  ∂ EC θ; θ
d C
QiiF =  . − (37)
dt  ∂ qi  ∂ qi
Symmetry 2022, 14, 81 21 of 30

According to [41], the kinetic energy for the 6R considered structure is determined with:
 . 6  
1 1
EC θ, θ = ∑ 2
T i
· Mi · i v Ci · v Ci + · i ω iT · i I i∗ · i ω i
2
(38)
i =1

where Mi is the mass, i v Ci and i ω i represent the linear and angular velocity of the mass
center for each joint i I i∗ inertial axial-centrifugal tensor of kinetic joint (i ), according to the
frame applied in the joint’s mass center.
The mass parameters for the CardioVR-ReTone structure were established using
SolidWorks software application, as the results from Figure 8 highlight for joints J4 and J5.

Figure 8. Mass parameters for the CardioVR-ReTone exoskeleton. (a) Joint 4—mass properties.
(b) Joint 5—mass properties.

The generalized gravitational forces are defined as:

T
Qig θ = 0 J i θ 0



· F xi θ , (39)

T
where 0 J i θ represents the line (i ) of the transposed of Jacobian matrix defined in (28),
and 0 F xi θ is the resultant force-moment vector of gravitational loads for [41].

The generalized handling forces are expressed as [49]:

QiSU (θ ) = 0 J iT (θ ) · 0 F X (θ ) , i = 4, 5. (40)

where 0 F X (θ ) is the vector of the resultant force-load handling moment.

The precise determination of these forces and moments plays an essential role in the
correct design of a kinematic axis. Thus, taking into account the mode and kinematic chain
of motion transmission of each motor coupling [50], substituting the expressions of the
generalized gravitational, manipulation, inertia, previously obtained forces in the definition
expression (35), the generalized driving forces, in analytical form on the output shafts of
the drive systems, have the following final form:
..
Q4m = 0.20267588 · q4jk + 0.05685865 (41)
..
Q5m = 0.429281378 · q5jk − 0.0009685371 (42)
The previously obtained relationships, (41) and (42), are in accordance with the kine-
matic structure of the CardioVR-ReTone robotic exoskeleton, but without considering
the frictions. As a central observation, it should be remembered that generally, through
the correct and real determination of the driving moments, the motors can be rigorously
Symmetry 2022, 14, 81 22 of 30

dimensioned for the actuation of a kinematic axle or the associated braking systems, thus
avoiding critical situations that can lead to damage to the mechanical structure.
The generalized variables that express the movements of the mechanical system joints
will be replaced by polynomial time functions according to the working process of the
robotic exoskeleton, as described in Section 3.2.

3.2. Working Process of Kinetic Joints Using (3n) − type Polynomial Functions
The functioning of the kinetic links i = 4, 5 on j = 1 → 3 are considered working se-
quences, with each sequence being divided into three segments, hence the result k = 1 → 6 .
In order to highlight the variation of the kinematic parameters, as well as the driv-
ing moments, according to [44,50,51], each sequence of the process is analyzed by in-
terpolating the motion trajectory in the configurations space on each working sequence,
with cubic spline functions and complying with the restrictions imposed by the driving
and control systems of the structure. The establishment of interpolation cubic splines of
(3n) − type polynomial functions is based on the generation of linear time functions for
the generalized accelerations of each driving joint of the robot. According to the work-
ing process, the motion trajectory in the configuration space must pass through all of
the points corresponding to the moments [τi (i = 0 → n)]. In addition, according to the
initial conditions, the movement path must provide a restrictive control over the position,
speed, and acceleration at times τ0 and τn , respectively, with continuity of speed and
acceleration at [τk (k = 1 → n − 1)]. Therefore, the interpolation of each segment of the
trajectory k = 1 → 6 is done using cubic spline functions. For their determination, a linear
function, with respect to time, is generated for the generalized accelerations of each driving
joint [50,51], as follows:

.. τi − τ .. τ − τi−1 ..
q ji (τ ) = · q ji (τi−1 ) + · q ji (τi ) (43)
ti ti

where ti = τi − τi−1 is the time necessary to run the segment (i = 1 → 3) of the trajectory.
By integrating the differential Equation (43), the following functions are obtained:

. (τi − τ )2 .. (τ − τi−1 )2 ..
q ji (τ ) = − · q ji−1 + · q ji + a ji1 ; (44)
2 · ti 2 · ti

(τi − τ )3 .. (τ − τi−1 )3 ..
q ji (τ ) = · q ji−1 + · q ji + a ji1 · τ + a ji2 (45)
6 · ti 6 · ti
For the study of the working process, with the help of the concepts presented above,
the numerical values imposed for the coordinates and the running times of the intervals
are presented in Table 8.

Table 8. Coordinates and the running times for joints J4 and J5 from CardioVR-ReTone.

Joint
Configuration Coordinates Values Duration
Joint i Seq. j = 4,5 Rotation Reported to Time τjk hsi
k = 10→15 qijk [rad] ti hsi
Previous Position (◦ )
9 0 0 0 6.77832
4 4 10 60 1.047197551 0.8334 7.61172
11 120 2.094395102 0.8334 8.44512
12 180 3.141592654 0.8334 9.27852
12 0 0 0 9.27852
5 5 13 75 1.308996939 1.04175 10.32027
14 150 2.617993878 1.04175 11.36202
15 225 3.926990817 1.04175 12.40377
Symmetry 2022, 14, 81 23 of 30

Based on the previous considerations, the expressions that characterize the positions,
velocities, and accelerations on each (k = 1÷3) segment of the sequence are presented in
Table 9.

Table 9. Positions, velocities, and accelerations for joints J4 and J5.

Expressions for Generalized Positions, Velocities, and Accelerations

Sequence Interval . ..
j = 4,5 Coordinate qijk qijk qijk
k = 1→3
[rad] [rad/s] [rad/s2 ]
1 0.912 · (τ − 6.78)3 − 1.05 · 10−19 · τ + 7.4 · 10−19 2.74 · (τ − 6.78)2 − 1.085 · 10−19 5.47 · τ − 37.09
4 2 q4 1.82 · τ 3 + 43.77 · τ 2 − 348.47 · τ + 918.73 −5.46 · τ 2 + 87.54 · τ − 348.48 87.54 − 10.91 · τ
3 2
3 0.89 · (τ − 9.28) + 3.14 2.69 · (τ − 9.28) 5.37 · τ − 49.86
3
0.5808109·(τ − 9.28) − 5.421010 · 10−20 · τ+
1 1.743 · (τ − 9.28)2 − 5.5 · 10−20 3.49 · τ − 32.34
+5.03 · 10−19
5 q5
2 −1.162 · τ 3 + 37.77 · τ 2 − 406.665 · τ + 1450.95 −3.485 · τ 2 + 75.55 · τ − 406.667 −6.97 · τ +75.56
3 2
3 2.169 · 10−19 · τ + 0.581 · (τ − 12.4) + 3.92 1.743 · (τ − 12.4) + 2.17 · 10−19 3.49 · τ − 43.22

Taking into account the expressions contained in Table 5, on the basis of expressions
(40) and (41), the variations of the driving moments are illustrated in Figures 9 and 10.

Figure 9. Variation of the generalized coordinates, speeds, and accelerations for J4.

Figure 10. Variation of the generalized coordinates, speeds, and accelerations for J5.

The run time for each trajectory segment, the duration of the segment, and the coor-
dinate values at the beginning and end of the sequence were used as the input data, and
the expressions of the generalized coordinates, speeds, and accelerations are presented in
Figure 11.
Symmetry 2022, 14, 81 24 of 30

Figure 11. Variation expressions of the driving moments for J4, and J5.

To evaluate, from an ergonomic point of view, the developed CardioVR-ReTone robotic

exoskeleton, the rapid upper limb assessment (RULA) method was used. The RULA
method was used to “rapidly” evaluate the exposure of the cardiac patient to ergonomic
risk factors associated with upper extremity musculoskeletal disorders (MSD).
The CardioVR-ReTone exoskeleton structure was analyzed with the RULA tool inte-
grated within CATIA software. The analysis presupposed a systematic process to evaluate
the required body posture, force, and repetition for each of the necessary rehabilitation
movements being evaluated (Figure 12). Considering the specificity of the cardiac reha-
bilitation process for cardiac patients following open heart surgery, in the first phase they
must recover the range of motion in the upper limbs, which should happen within the first
60 days after surgery. The recovery of the upper limbs’ range of motion is done gradually,
starting with light exercises with limited ROM to ample movements with full/almost
full ROM.

Figure 12. Evaluating the exposure of the cardiac patient to ergonomic risk factors—RULA method.

By analyzing the obtained results using the RULA method, as highlighted in Figure 12,
we identified the muscular effort associated with the necessary rehabilitation postures
and excessive forces while performing the necessary exercises, and which of the exercises
contribute to muscle fatigue. Thus, it can be said that none of the movements and positions
overload the muscles and joints excessively. The values obtained, as can be seen in Figure 12,
Symmetry 2022, 14, 81 25 of 30

are in the range of 1–2: neglectable risk, no action needed; or 2–4: low risk, change
may be needed. Moreover, the results obtained reflect the fact that both the 3D model
(mechanical design) and the kinematic and dynamic model (actuating and control system)
are technically feasible and those exercises necessary for CR can be performed with the
help of the exoskeleton without much difficulty.
Regarding the mechanical and kinematical CardioVR-ReTone movement possibilities,
Figure 13 shows the robot’s maximum working space. Within the working space, some
movements will be restricted with the kinematic model (singularities) and medical protocol
of CR. As can be seen, the CardioVR-ReTone is capable of performing the necessary CR
movements in both the joints and TCP.

Figure 13. The maximum working space of the CardioVR-ReTone exoskeleton.

Based on the results obtained in this research and highlighted in this article, we moved
on to the physical manufacturing of the CardioVR-ReTone exoskeleton. Figure 14 highlights
the final phase of assembling the mechanical and electrical structure of the exoskeleton.
CardioVR-ReTone’s prototype is planned for completion and initial lab testing by the end
of 2021. The exoskeleton will then be tested on volunteers, and a series of tweaks will be
made depending on the volunteers’ feedback, before being tested on cardiac patients.

Figure 14. The mechanical structure of the CardioVR-ReTone robotic exoskeleton.

Symmetry 2022, 14, 81 26 of 30

4. Discussion
A new robotic exoskeleton with 12 DoFs for the upper limbs used in the rehabilitation
of cardiac patients after open heart surgery was introduced in this paper. Our system uses
six motors on each arm. Thus, in this article, the CardioVR-ReTone robotic exoskeleton,
being symmetric, was considered a 6R serial structure robot for each arm. Section 2.2
presented the geometrical and kinematic model of the right part of the exoskeleton—6R
serial structure. The direct and inverse geometry equations for the mentioned structure
were established using the matrix of locating algorithm, considering the minimum number
of geometric or mechanical restrictions. The operational kinematic parameters expressing
the movement of the end effector in Cartesian space were defined by the iterative algorithm.
The kinematic control functions of the serial structure of type 6R were determined based on
the inverse of the Jacobian matrix. As a major point to consider, the algorithmizing of the
geometric and kinematic modelling of the CardioVR-ReTone robotic exoskeleton structures
had some major advantages, as a simple geometric visualization of the characteristics of
the different parameters considered the high degree of generalization, regardless of the
complexity of the mechanical systems involved.
The dynamic control functions were determined using the fundamental notion as
kinetic energy in the advanced mechanics of the mechanical systems. Starting from the
differential principles of the analytical mechanics, specific to the holonomic mechanical
systems, in which the serial structure of type 6R is inscribed and using the algorithm of the
generalized forces in dynamics, the generalized driving forces were directly determined
using analytical calculations. Further, the generalized variables that express the movement
of the mechanical system can be replaced by polynomial functions of time according to the
working process in which the serial robot structure is integrated. The polynomial functions
of interpolation were determined as real variables for the description of the motion process.
The expressions were substituted into the dynamic equations that led to the establishment
of the variation laws of the driving moments.
Based on the kinematic and dynamic model, the control architecture was developed,
and the electrical motors were chosen. Thus, the EC60 flat BLDC electric motor, coupled
with harmonic drives, was chosen due to its good torque-to-weight ratio, compact size, low
backlash, and ease of control. Considering the calculated torque, and taking into account
the economic factors like discounts for more than three identical motors, the following
actuators were selected for each exoskeleton arm: two AC-CPU17-100, one AC-CPU20-100,
and three AC-CPU25-100. Every actuator has an incremental encoder.
The control architecture of the CardioVR-ReTone was developed to allow the deploy-
ment of several exoskeleton control modes: assistive, partially assistive, and resistive. An
industrial open controller manufactured by Siemens, CPU 1515SP PC2, was considered the
main control unit. The PLC had a CANOpen communication module, used to connect with
every actuator EPOS4 controller and several IOs modules used to connect the signals from
the exoskeleton joysticks and mechanical safety limit switches mounted on the exoskele-
ton’s structure. The mechanical safety limit switches were used to activate the safe torque
off functionality of the actuators in cases of malfunction, considering the positioning pro-
cess. The open controller operating system hosted the VR engine and Simulink. Simulink
ran the kinematic algorithms and sent the positioning, speed, and torque data to every joint
actuator controller. Feedback was received from every actuator controller and sent back to
the Simulink Kinematics algorithm. Surface EMGs placed on human arms sent information
about muscle activity using Bluetooth to the open controller, which fed the data to Simulink
for processing and torque adjustment. The architecture of the CardioVR-ReTone operating
platform was based on a distributed control approach; the ROS (Robot Operating System)
platform is currently considered the first option for implementing control-related functions.
The early rehabilitation process of a patient after open-heart surgery is often long,
painful, and lacks challenges and fun, especially in an unfavorable context like difficulties
reaching a rehabilitation centre. Therefore, we envisaged extrinsic motivation mechanisms.
These mechanisms were included in two game components—one to accompany the patient
Symmetry 2022, 14, 81 27 of 30

along their rehab journey, and one as a simple game that can be played while performing the
exercises of a specific rehab level. The movements that the patient makes can be “translated”
into virtual keypress events, and thus, via the “exoskeleton-as-keyboard” module, even
external simple games can be used as motivational mechanisms. The entire rehabilitation
process lifecycle, starting immediately after surgery, continuing at the patient’s home,
and ending once the patient is completely recovered, drove the design of the interaction
scenarios described in Section 2.5. Similarly, the architecture of the CardioVR operating
software was built to support the interaction scenarios of each user role (patient, therapist,
and therapist assistant).
The article presents the CardioVR-ReTone exoskeleton, that has not been implemented
in clinical practice yet, and might be, to our knowledge, the first model of its kind used
for early CR after sternotomy. The CardioVR-ReTone exoskeleton successfully passed
the software simulation tests. Before its implementation in clinical practice, the robotic
system needs to pass laboratory tests with healthy volunteers. Further studies should be
conducted to balance the efficacity, advantages, and disadvantages of the CardioVR-ReTone
exoskeleton, as well as its sustainability and utility in clinical practice. Our entire medical
testing protocol is described by our group in [39]. The efficacity of the system will be also
compared to usual CR, and will be assessed using several variables such as: quality of life;
sternal stability; upper arm skeletal muscle function assessment; pain intensity; exercise
protocol; cardiac response to exercise; and compliance/adherence dropout rates.

5. Conclusions
In this paper, the electromechanical design (geometric, kinematic, and dynamic model),
control architecture, and VR software module of the CardioVR-ReTone robotic exoskeleton
was presented. The CardioVR-ReTone robotic exoskeleton consists of two serial robotic
arms, each with six DoFs, and with mirrored joints. The geometric, kinematic, and dynamic
model presented in this article was developed for the right arm of the exoskeleton. From
the geometric point of view, the two robotic arms are identical and the kinematic main
difference between the right and left robotic arms is the rotation direction of certain joints.
The matrix of locating algorithm was used to determine the direct and inverse geome-
try equations for the stated construction, while taking into account the minimum number
of geometric or mechanical constraints. The iterative algorithm defined the operational
kinematic parameters representing the movement of the end effector in Cartesian space.
The kinematic control functions of the serial structure of type 6R were determined based on
the inverse of the Jacobian matrix. The dynamic control functions were calculated utilizing
kinetic energy, a key concept in advanced mechanical systems mechanics. The generalized
driving forces were directly computed using the algorithm of the generalized forces in
dynamics and the differential principles of analytical mechanics, which are peculiar to
holonomic mechanical systems in which the serial structure of type 6R is inscribed.
Developing a robotic exoskeleton for upper limbs that can support a cardiac patient
following open-heart surgery during the early rehabilitation process necessitates a deep
understanding of the kinematics and dynamics of the human upper limbs. The CardioVR-
ReTone exoskeleton was devised following debates and discussions with cardiologists. The
six-DoF robotic exoskeleton design relied significantly on the results obtained from these
interactions to offer the necessary engineering specifications.
The design of the robotic exoskeleton included the proximal positioning of the har-
monic drive’s motors, and open mechanical human–machine interfaces. High stiffness link-
ages and back-drivable transmissions without backlash are key features of the CardioVR-
ReTone exoskeleton. The proposed exoskeleton structure does not use stiff elements for
coupling the human arm with the exoskeleton articulation; instead, it relies on the rotation
center to ensure optimal patient elbow joint articulation. This has a big impact on the
comfort level of the human–robot interaction.
Symmetry 2022, 14, 81 28 of 30

In terms of future development, the intention is to control the CardioVR-ReTone

robotic exoskeleton with the help of a brain-controlled helmet. So far, we have managed to
start an electric motor with the help of an Emotiv-Epoc+ 14-channel wireless EEG headset.

Author Contributions: Conceptualization, formal analysis, and funding acquisition, B.M.; methodol-
ogy, C.S., M.M. (Mircea Murar), and M.F.; software, M.F. and R.C.; validation, formal analysis, C.S.
and C.N.; investigation, M.M. (Mircea Murar), M.M. (Mihaela Mocan); resources, data curation, C.S.
and B.M; writing—original draft preparation, B.M.; writing—review and editing, M.M. (Mihaela
Mocan); visualization, validation, M.F.; supervision, B.M.; project administration, B.M. All authors
have read and agreed to the published version of the manuscript.
Funding: This research was funded by a grant of the Romanian Ministry of Education and Research,
CCCDI—UEFISCDI, project number PN III-P2-2.1-PED-2019-1057, within PNCDI III.
Institutional Review Board Statement: The study will be conducted according to the guidelines of
the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of
University of Medicine and Pharmacy Iuliu Hatieganu Cluj-Napoca Romania (registration number
350/2.10.2019).
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author. The data are not publicly available due to privacy issues.
Acknowledgments: This work was supported by a grant from the Romanian Ministry of Education
and Research, CCCDI—UEFISCDI, project number PN III-P2-2.1-PED-2019-1057, within PNCDI III.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or
in the decision to publish the results.

References
1. Ruiz-Olaya, A.F.; Lopez-Delis, A.; da Rocha, A.F. Upper and Lower Extremity Exoskeletons. In Handbook of Biomechatronics;
Segil, J., Ed.; Elsevier Inc.: London, UK, 2018.
2. Mocan, M.; Răhăian, R.; Mocan, B.; Blaga, S.N. Multi-Marker Evaluation of the Cardio-Metabolic Risk. Atherosclerosis 2017, 263,
E201–E202. [CrossRef]
3. Hamaya, M.; Matsubara, T.; Noda, T.; Teramae, T.; Morimoto, J. Learning assistive strategies for exoskeleton robots from
user-robot physical interaction. Pattern Recognit. Lett. 2017, 99, 67–76. [CrossRef]
4. Martínez, G.M.C.; Zuniga Avilés, L.A. Design Methodology for Rehabilitation Robots: Application in an Exoskeleton for Upper
Limb Rehabilitation. Appl. Sci. 2020, 10, 5459. [CrossRef]
5. Mocan, M.; Mocan, B. Cardiac rehabilitation for older patients with cardiovascular pathology using robotic systems—A survey.
Balneo Res. J. 2019, 10, 33–36. [CrossRef]
6. Schabron, B.; Desai, J.; Yihun, Y. Wheelchair-Mounted Upper Limb Robotic Exoskeleton with Adaptive Controller for Activities
of Daily Living. Sensors 2021, 21, 5738. [CrossRef]
7. Mocan, M.; Chiorescu, R.; Banc, O.N.; Mocan, B.; Anton, F.; Stoia, M.; Farcas, A.D. Cardiac rehabilitation protocols outcome in
frail patients undergoing transcatheter aortic valve implantation. BALNEO Res. J. 2018, 9, 401–405. [CrossRef]
8. Wu, Q.; Wu, H. Development, Dynamic Modeling, and Multi-Modal Control of a Therapeutic Exoskeleton for Upper Limb
Rehabilitation Training. Sensors 2018, 18, 3611. [CrossRef]
9. Galofaro, E.; D’Antonio, E.; Patané, F.; Casadio, M.; Masia, L. Three-Dimensional Assessment of Upper Limb Proprioception via a
Wearable Exoskeleton. Appl. Sci. 2021, 11, 2615. [CrossRef]
10. Liu, C.; Liang, H.; Ueda, N.; Li, P.; Fujimoto, Y.; Zhu, C. Functional Evaluation of a Force Sensor-Controlled Upper-Limb
Power-Assisted Exoskeleton with High Backdrivability. Sensors 2020, 20, 6379. [CrossRef]
11. Rosen, J.; Brand, M.; Fuchs, M.B.; Arcan, M. A myosignal-based powered exoskeleton system. IEEE Trans. Syst. Man Cybern. Part
A Syst. Hum. 2001, 31, 210–222. [CrossRef]
12. Fanin, C.; Gallina, P.; Rossi, A.; Zanatta, U.; Masiero, S. NeRebot: A wire-based robot for neurorehabilitation. In Proceedings of
the 9th International Conference on Rehabilitation Robotics, Taejon, Korea, 23–25 April 2003; pp. 23–27.
13. Nef, T.; Mihelj, M.; Kiefer, G.; Perndl, C.; Müller, R.; Riener, R. ARMin—Exoskeleton for arm therapy in stroke patients. In
Proceedings of the ICORR’07, 2007 IEEE 10th International Conference on Rehabilitation Robotics, Noordwijk, The Netherlands,
13–15 June 2007; pp. 68–74. [CrossRef]
Symmetry 2022, 14, 81 29 of 30

14. Garrec, P.; Friconneau, J.P.; Méasson, Y.; Perrot, Y. ABLE, an innovative transparent exoskeleton for the upper-limb. In Proceed-
ings of the 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems, Nice, France, 22–26 September 2008;
pp. 1483–1488. [CrossRef]
15. Brackbill, E.A.; Mao, Y.; Agrawal, S.K.; Annapragada, M.; Dubey, V.N. Dynamics and control of a 4-dof wearable cable-driven
upper arm exoskeleton. In Proceedings of the 2009 IEEE International Conference on Robotics and Automation, Kobe, Japan,
12–17 May 2009; pp. 2300–2305. [CrossRef]
16. Rahman, M.H.; Kittel-Ouimet, T.; Saad, M.; Kenné, J.-P.; Archambault, P.S. Development and control of a robotic exoskeleton for
shoulder, elbow and forearm movement assistance. Appl. Bionics Biomech. 2012, 9, 275–292. [CrossRef]
17. Stienen, A.H.A.; Hekman, E.E.G.; Prange, G.B.; Jannink, M.J.A.; Aalsma, A.M.M.; Van Der Helm, F.C.T.; Van Der Kooij, H.
Dampace: Design of an exoskeleton for force-coordination training in upper-extremity rehabilitation. J. Med. Devices 2009, 3,
031003. [CrossRef]
18. Naidu, D.; Stopforth, R.; Bright, G.; Davrajh, S. A Portable Passive Physiotherapeutic Exoskeleton. Int. J. Adv. Robot. Syst. 2012, 9,
137. [CrossRef]
19. Pirondini, E.; Coscia, M.; Marcheschi, S.; Roas, G.; Salsedo, F.; Frisoli, A.; Bergamasco, M.; Micera, S. Evaluation of a New
Exoskeleton for Upper Limb Post-stroke Neuro-rehabilitation: Preliminary Results. Biosyst. Biorobotics 2014, 7, 637–645. [CrossRef]
20. Zhou, L.; Bai, S.; Andersen, M.S.; Rasmussen, J. Modeling and design of a spring-loaded, cable-driven, wearable exoskeleton for
the upper extremity. Model. Identif. Control 2015, 36, 167–177. [CrossRef]
21. Islam, M.R.; Assad-Uz-Zaman, M.; Brahmi, B.; Bouteraa, Y.; Wang, I.; Rahman, M.H. Design and Development of an Upper Limb
Rehabilitative Robot with Dual Functionality. Micromachines 2021, 12, 870. [CrossRef]
22. Gull, M.A.; Thoegersen, M.; Bengtson, S.; Mohammadi, M.; Struijk, L.A.; Moeslund, T.; Bak, T.; Bai, S. A 4-DOF Upper Limb
Exoskeleton for Physical Assistance: Design, Modeling, Control and Performance Evaluation. Appl. Sci. 2021, 11, 5865. [CrossRef]
23. Huamanchahua, D.; Vargas-Martinez, A.; Ramirez-Mendoza, R. Kinematic of the Position and Orientation Synchronization of the
Posture of a n DoF Upper-Limb Exoskeleton with a Virtual Object in an Immersive Virtual Reality Environment. Electronics 2021,
10, 1069. [CrossRef]
24. Chen, C.-T.; Lien, W.-Y.; Chen, C.-T.; Wu, Y.-C. Implementation of an Upper-Limb Exoskeleton Robot Driven by Pneumatic
Muscle Actuators for Rehabilitation. Actuators 2020, 9, 106. [CrossRef]
25. Zhao, Y.; Liang, C.; Gu, Z.; Zheng, Y.; Wu, Q. A New Design Scheme for Intelligent Upper Limb Rehabilitation Training Robot.
Int. J. Environ. Res. Public Health 2020, 17, 2948. [CrossRef]
26. Buongiorno, D.; Cascarano, G.D.; Camardella, C.; De Feudis, I.; Frisoli, A.; Bevilacqua, V. Task-Oriented Muscle Synergy
Extraction Using An Autoencoder-Based Neural Model. Information 2020, 11, 219. [CrossRef]
27. Badesa, F.J.; Diez, J.A.; Catalan, J.M.; Trigili, E.; Cordella, F.; Nann, M.; Crea, S.; Soekadar, S.R.; Zollo, L.; Vitiello, N.; et al.
Physiological Responses During Hybrid BNCI Control of an Upper-Limb Exoskeleton. Sensors 2019, 19, 4931. [CrossRef]
[PubMed]
28. Chen, W.; Li, Z.; Cui, X.; Zhang, J.; Bai, S. Mechanical design and kinematic modeling of a cable-driven arm exoskeleton
incorporating inaccurate human limb anthropomorphic parameters. Sensors 2019, 19, 4461. [CrossRef] [PubMed]
29. Chaparro-Rico, B.D.M.; Cafolla, D.; Ceccarelli, M.; Castillo-Castaneda, E. NURSE-2 DoF Device for Arm Motion Guidance:
Kinematic, Dynamic, and FEM Analysis. Appl. Sci. 2020, 10, 2139. [CrossRef]
30. Bond, S.; Laddu, D.R.; Ozemek, C.; Lavie, C.J.; Arena, R. Exergaming and Virtual Reality for Health: Implications for Cardiac
Rehabilitation. Curr. Probl. Cardiol. 2021, 46, 100472. [CrossRef] [PubMed]
31. The Top 15 Examples of Gamification in Healthcare. Available online: https://medicalfuturist.com/top-examples-of-
gamification-in-healthcare/ (accessed on 1 October 2021).
32. Dithmer, M.; Rasmussen, J.O.; Grönvall, E.; Spindler, H.; Hansen, J.S.; Nielsen, G.D.; Sorensen, S.B.; Dinesen, B.I. “The Heart
Game”: Using Gamification as Part of a Telerehabilitation Program for Heart Patients. Games Health J. 2016, 5, 27–33. [CrossRef]
33. Ingadottir, B.; Jaarsma, T.; Klompstra, L.; Aidemark, J.; Askenäs, L.; Bahat, Y.; Ben Gal, O.; Berglund, A.; Berglund, E.; Höchsmann,
C.; et al. Let the games begin: Serious games in prevention and rehabilitation to improve outcomes in patients with cardiovascular
disease. Eur. J. Cardiovasc. Nurs. 2020, 19, 558–560. [CrossRef]
34. Stuart, A.G. Exercise as therapy in congenital heart disease—A gamification approach. Prog. Pediatr. Cardiol. 2014, 38, 37–44.
[CrossRef]
35. Lee, J.G.; Choi, Y.; López, J. A Review and Evaluation of Smartphone Applications to Support Virtual Cardiac Rehabilitation. J.
Am. Coll. Cardiol. 2020, 75, 3644. [CrossRef]
36. Ambrosetti, M.; Abreu, A.; Corrà, U.; Davos, C.H.; Hansen, D.; Frederix, I.; Iliou, M.C.; E Pedretti, R.F.; Schmid, J.-P.; Vigorito, C.;
et al. Secondary prevention through comprehensive cardiovascular rehabilitation: From knowledge to implementation. 2020
update. A position paper from the Secondary Prevention and Rehabilitation Section of the European Association of Preventive
Cardiology. Eur. J. Prev. Cardiol. 2020, 28, 460–495. [CrossRef]
37. Doyle, M.P.; Indraratna, P.; Tardo, D.T.; Peeceeyen, S.C.S.; Peoples, G.E. Safety and efficacy of aerobic exercise commenced early
after cardiac surgery: A systematic review and meta-analysis. Eur. J. Prev. Cardiol. 2019, 26, 36–45. [CrossRef] [PubMed]
38. Katijjahbe, M.A.; Granger, C.L.; Denehy, L.; Royse, A.; Royse, C.; Bates, R.; Logie, S.; Ayub, A.N.; Clarke, S.; El-Ansary, D.
Standard restrictive sternal precautions and modified sternal precautions had similar effects in people after cardiac surgery via
median sternotomy (“SMART” Trial): A randomised trial. J. Physiother. 2018, 64, 97–106. [CrossRef]
Symmetry 2022, 14, 81 30 of 30

39. Mocan, M.; Vlaicu, S.I.; Farcas, , A.D.; Feier, H.; Dragan, S.; Mocan, B. Cardiac Rehabilitation Early after Sternotomy Using New
Assistive VR-Enhanced Robotic Exoskeleton—Study Protocol for a Randomised Controlled Trial. Int. J. Environ. Res. Public Health
2021, 18, 11922. [CrossRef] [PubMed]
40. Negrean, I.; Kacso, K.S.; Schonstein, C.; Duca, A. Mecanică: Teorie şi Aplicaţii; UT Press: Cluj-Napoca, Romania, 2012.
41. Negrean, I.; Duca, A.; Negrean, C.; Kacso, K.S. Mecanica Avansată în Robotică; UT Press: Cluj-Napoca, Romania, 2008.
42. Schonstein, C.; Negrean, I.; Kacso, K. Using of Polynomial Functions in Modelling of the Working Process of Mobile Robot
RmITA. Acta Tech. Napoc. Ser. Appl. Math. Mech. Eng. 2014, 57, 195–200.
43. Negrean, I.; Kacso, K.; Schonstein, C.; Duca, A.; Rusu, F.; Cristea, F.; Haragas, S. New Formulations on Acceleration Energies in
Analytical Dynamics. Appl. Mech. Mater. 2016, 823, 43–48. [CrossRef]
44. Negrean, I.; Schonstein, C.; Duca, A. Dynamic Model for the Mobile Robot RmITA. Acta Tech. Napoc. Ser. Appl. Math. Mech. Eng.
2014, 57, 159–164.
45. The Robot Operating System (ROS). Available online: https://www.ros.org/ (accessed on 10 October 2021).
46. Çubukçu, B.; Yüzgeç, U.; Zileli, R.; Zileli, A. Reliability and validity analyzes of Kinect V2 based measurement system for shoulder
motions. Med. Eng. Phys. 2020, 76, 20–31. [CrossRef] [PubMed]
47. Chaparro-Rico, B.D.M.; Cafolla, D.; Castillo-Castaneda, E.; Ceccarelli, M. Design of arm exercises for rehabilitation assistance. J.
Eng. Res. 2020, 8, 203–218. [CrossRef]
48. van Diest, M.; Stegenga, J.; Wörtche, H.J.; Postema, K.; Verkerke, G.J.; Lamoth, C.J.C. Suitability of Kinect for measuring whole
body movement patterns during exergaming. J. Biomech. 2014, 47, 2925–2932. [CrossRef]
49. Schonstein, C. Kinematic Control Functions for a Serial Robot Structure Based on the Time Derivative Jacobian Matrix. Acta Tech.
Napoc. 2018, 61, 219–224.
50. Negrean, I. New Formulations in Analytical Dynamics of Systems. Acta Tech. Napoc. Ser. Appl. Math. Mech. Eng. 2017, 60, 49–56.
51. Negrean, I.; Schonstein, C.; Kacso, K.; Duca, A. Formulations in advanced dynamics of mechanical systems. Mech. Mach. Sci.
2014, 17, 185–195. [CrossRef]