Original Article

Proc IMechE Part H:

J Engineering in Medicine
2022, Vol. 236(5) 697–710
Human-robot-robot cooperative Ó IMechE 2022
Article reuse guidelines:
control using positioning robot and sagepub.com/journals-permissions
DOI: 10.1177/09544119221083140

1-DOF traction device for robot- journals.sagepub.com/home/pih

assisted fracture reduction system

Woo Young Kim1, Sanghyun Joung2, Il Hyung Park3,

Jong-Oh Park1 and Seong Young Ko1

Abstract
While performing musculoskeletal long bone fracture reduction surgery, assistant surgeons can often suffer from physical fati-
gue as they provide resistance against the tension from surrounding muscles pulling on the patient’s broken bones. These days,
robotic systems are being actively developed to mitigate this physical workload by realigning and holding these fractured bones
for surgeons. This has led to one consortium proposing the development of a robot-assisted fracture reduction system consist-
ing of a 6-DOF positioning robot along with a 1-DOF traction device. With the introduction of the 1-DOF traction device, the
positioning robot does not have to fight these contraction forces so can be compact improving its maneuverability and overall
convenience; however, considering surgeon-robot interactions, this approach adds the requirement of controlling two different
types of robots simultaneously. As such, an advanced cooperative control methodology is required to control the proposed
bone fracture reduction robot system. In this paper, a human-robot-robot cooperative control (HRRCC) scheme is proposed
for collaboration between the surgeon, the positioning robot, and the traction device. First, the mathematical background of
this HRRCC scheme is provided. Next, we describe a series of experiments that show how the proposed scheme facilitates a
reduction in the load placed on the positioning robot from strong muscular contraction forces making it possible to conduct
fracture reduction procedures more safely despite the muscular forces.

Keywords
Admittance control, robot-assisted fracture reduction, human-robot-robot cooperative control, positioning robot, trac-
tion device

Date received: 25 October 2018; accepted: 24 January 2022

Introduction operating surgeon align the fractured bones. Since the

muscles surrounding the bones will strongly contract in
Minimally invasive osteosynthesis, especially closed these situations, considerable force is required to posi-
intramedullary nailing, has recently become the preferred tion then hold the bones in place. It has been reported
method for long bone fracture reduction surgery due to that the maximum force and torque exerted by thigh
the successful surgical outcomes it produces; however, sig- muscle tension in adult males can measure up to 411 N
nificant disadvantages to this approach have been
reported, including the physically difficult tasks that sur-
geons must perform, potential misalignment of fracture 1
School of Mechanical Engineering, Chonnam National University,
segments, and the large amounts of radiation exposure Gwangju, South Korea
2
experienced by patients as well as operating staff. Medical Device and Robot Institute of Park, Kyungpook National
University, Daegu, South Korea
3
School of Medicine, Department of Orthopedic Surgery, Kyungpook
National University, Daegu, South Korea
Necessity for robot-assisted fracture reduction
systems Corresponding author:
Seong Young Ko, School of Mechanical Engineering, Chonnam National
During femur fracture reduction surgery, several assis- University, 77 Yongbong-ro, Buk-gu, Gwangju 61186, South Korea.
tant surgeons pull and hold a patient’s leg to help the Email: [email protected]
698 Proc IMechE Part H: J Engineering in Medicine 236(5)

and 74 Nm, respectively.1 Recently, in order to mitigate research group has previously developed a RAFR sys-
the physical burden of assistant surgeons during these tem in which one serial type positioning robot is used
procedures, and to reduce X-ray radiation exposure for solely to provide force feedback and force scale func-
patients and operating staff, robot-assisted fracture tionality this system operates based on the assumptions
reduction (RAFR) systems equipped with 3-dimen- that the positioning robot has a sufficient load capacity
sional (3D) medical navigation systems have been and that two F/T sensors are available.10 Recently,
developed by several research groups.2–9 Park et al.13,14 proposed a new approach where a
single-axis motorized traction device is used to assist
with the strong muscular contraction force exerted
Literature review along the bone’s axial axis alongside a serial robot.
Several research groups have developed RAFR systems This combination is promising because the largest
using powerful robots to mitigate surgeons’ physical forces are required along the axial axis of the bone,
fatigue. Westpal et al.5 proposed a robot-assisted long while the forces in the other directions are lower.1,15,16
bone fracture reduction system that employs a six-axis Thus, for that system, the traction device manages the
articulated industrial robot (Stäubli TX-90) and a 3D strong muscular forces acting along the bone’s axial
image navigation system, which is telemanipulated axis, while the serial robot manages the remaining
using a joystick. Tokyo University in Japan6 proposed forces along other axes. The small serial robot has been
a RAFR system (FRAC-Robo) equipped with a six- named the ‘‘positioning robot.’’ Employing the separate
axis serial robot and a 3D image navigation system. traction device helps reduce the maximum payload on
This robot can be connected to a patient’s foot to move the positioning robot and improves the surgeon’s man-
the distal bone fragment while being interactively con- euverability. However, since two different types of
trolled using one force/torque (F/T) sensor. Palmerston robots are used together, a cooperative control scheme
North Hospital in New Zealand7 proposed a RAFR is required. If a positioning robot and a traction device
system that consisted of a 6-degree-of-freedom (DOF) move independently, the traction force produced by the
parallel robot and a manual reduction table. The traction device may create additional loads for the posi-
University of Hong Kong utilized a specially designed tioning robot to deal with. To conduct fracture reduc-
parallel robot in their long bone RAFR system.8 tion surgery based on a surgeon’s commands like one
Dagnino et al.9 developed a RAFR system that consists integrated system, an efficient control structure must
of a 6-DOF parallel robot and a 4-DOF carrier plat- be developed.
form. The above-mentioned RAFR systems are mainly Thus, for the cooperative control of both the posi-
manipulated by surgeons using teleoperation to con- tioning robot and the traction device, where the two
duct fracture reduction tasks.5–9 For some systems, robot systems are connected to the fractured bone phy-
experimental results from situations where the muscu- sically, a human-robot-robot cooperative control
lar contraction forces or soft tissues are considered (HRRCC) scheme is proposed. In detail, as the surgeon
explicitly have not been reported.5,8,9 Some systems are controls the positioning robot, the traction device
bulky,6 and some require use of additional structures, moves to minimize the load placed on the positioning
such as a carriage platform.7,9 In this study, teleopera- robot. It is expected that this scheme will help surgeons
tion along with manual interactive manipulation was conduct RAFR surgery while experiencing lower physi-
adopted as the intended control method so that sur- cal loads and stresses than before, in addition, the sys-
geons can intervene in the robotic surgery procedure tem will protect bone fragments from damage, such as
and manipulate the robot directly while checking a a secondary fracture caused by excessive forces from
patient’s status.10 the robots’ operations.
Moreover, during conventional fracture reduction In the field of industrial robots, initially coordina-
surgery, a manual traction device is often used. These tion of multiple robotic arms was developed using
manual traction devices may use weights to pull the coordinated control17 or by expanding hybrid position/
bone and pulleys to change the direction of that force, force control methods.18 Since then, adaptive con-
or they may use a manual winch system to pull and trol19,20 and neural network control21 techniques have
hold a broken bone in place. Generally, these devices been developed and applied for control of multiple
are fixed to the frame of a patient’s bed. In a similar robots. Motion synchronization techniques for multi-
way, fracture tables have also seen widespread use for axis robots have also been developed using cross-
assisting fracture reduction surgery. coupling approaches.22–24 It should be noted that the
For control, most surgical robot systems have a existing synchronization methods found in previous lit-
surgeon-robot cooperative control scheme so the sur- erature, like those mentioned above, are all related to
geon can manage surgical processes. Most systems motion control. Motion synchronization using force
include a sensor-based virtual guide11 and/or a force tracking control for human-robot collaboration has
feedback control to aid the user’s force-torque com- also been studied.25 In recent years, multiple robot
mands and environmental reaction forces.10,12 This arms have been studied for force control
Kim et al. 699

both assistance against contraction force and the reduc-

tion of contact forces. When bone fragments are placed
into contact with each other, the contact forces between
them can be reduced significantly, the force feedback
function of the proposed HRRCC then limits the dis-
placements of the positioning robot.

Human-robot-robot cooperative control

Dynamics modeling using simplified one-DOF
systems
In this section, the control structure and mathematical
background of the HRRCC scheme are described. A
conceptual layout of the two robot systems involved, a
Figure 1. Conceptual setup of a 6-DOF positioning robot and
a 1-DOF traction device for procedure on a fractured femur positioning robot and a traction device, are shown in
model with phantom muscles: (1) 6-DOF positioning robot, Figure 1. The fractured distal bone is firmly connected
(2) 1-DOF traction device, (3) Interaction F/T sensor, to the end of the 6-DOF positioning robot through a
(4) Operational F/T sensor, (5) Pin holder, (6) Distal bone round fixation jig, the positioning robot can then be
fixation jig, (7) Fractured femur with phantom muscle, moved to adjust the position and orientation of the dis-
(8) Proximal bone fixation jig. tal bone according to the surgeon’s maneuvering
motion to align it with the proximal bone. In order to
reduce the axial load on the positioning robot during
synchronization so that a specific robot arm can be the alignment process caused by the strong contractile
applied to a workpiece with the desired force.26 This forces from the muscles around the bone, a 1-DOF
work has proceeded to simulated experiments. traction device is connected in parallel with the posi-
tioning robot for traction of the distal bone. The proxi-
mal fractured bone is firmly fixed to a round bone
Contributions of this work fixation jig, and the fixation jig is fixed to the frame of
The first major contribution of this work is the pro- the surgical bed.
posed HRRCC scheme to control two different types In practice, it is expected that the HRRCC will be
of robots based on singular commands from the sur- applied to multiple-DOF systems; however, for simpli-
geon. To the best of the authors’ knowledge, this is the city, a one-DOF positioning robot and a one-DOF
first such approach for medical robots. Second, a traction device are considered and modeled in this
RAFR system was implemented in which a 1-DOF work. The overall system is represented as mass/spring/
traction device manages most of the muscular contrac- damper systems, as shown in Figure 2. This assumes
tion force while a bone fragment is being positioned that the muscles are passive leading to the choice of a
while a 6-DOF positioning robot manages the remain- spring-damper model. The positioning robot and the
ing forces on other axes. Third, it was experimentally traction device are both represented as linear mass-
verified that our single control structure can provide damper systems. Although the proximal bone is fixed

Figure 2. A simplified dynamics model including the positioning robot, traction device, human operator, and fractured bones with
muscles.
700 Proc IMechE Part H: J Engineering in Medicine 236(5)

The dynamics of the distal fractured bone, which

interacts with the positioning robot, the traction device
and the muscles, are modeled as follows:

me x€e + (bf + bc + be )x_ e + (kf + kc + ke )xe

ð3Þ
 bf x_ r  kf xr  bc x_ tr  kc xtr = 0

where me, be, and ke denote the mass coefficient of

the distal bone, the viscous coefficient and stiffness
(a) (b)
coefficient of the muscles between the bones,
respectively.
(1) to (3) can be transformed into s-domains using a
Laplace transform and can be rewritten as shown in (4)
to (6). Their block diagrams can be represented as
shown in Figure 3:

Gr (s) Gr (s)
x_ r = (t r + fm )+ x_ e
1 + Zf (s)Gr (s) 1 + Zf (s)Gr (s)
(c) ð4Þ

Figure 3. Block diagrams of (a) positioning robot and distal bone

where Gr (s) denotes the transfer function of the posi-
fixation jig derived from (4), (b) traction device and traction cable
derived from (5), and (c) fractured bone with muscle, distal bone tioning robot, that is, Gr (s) = 1=(mr s + br ), and Zf (s)
fixation jig, and traction cable derived from (6). denotes the impedance of the distal fixation jig, that is,
Zf (s) = bf + kf =s.
to the frame of the surgical bed through the proximal
Gtr (s) Z~c (s)Gtr (s)
fixation jig, for simplicity, the proximal bone is x_ tr = t tr + x_ e ð5Þ
assumed to be fixed directly to the ground. For each 1 + Zc (s)Gtr (s) 1 + Zc (s)Gtr (s)
distal bone fixation jig, the cable/pin holder and the
muscles are modeled as spring-damper systems. where Gtr (s) denotes the transfer function of the trac-
Finally, as an operator (the surgeon) manipulates the tion device, that is, Gtr (s) = 1=(mtr s + btr ), and Z~c (s)
positioning robot directly, he/she is considered to be denotes the impedance of the traction cable, that is,
connected to the positioning robot in series. Z~c (s) = bc + kc =s.
For this model, the dynamics of the positioning
robot and the distal bone fixation jig can be described Zf (s)Ge (s)
x_ e = x_ r
as follows: 1 + Zf (s)Ge (s) + Z~c (s)Ge (s)
ð6Þ
mr x€r +(br + bf )x_ r + kf xr  bf x_ e kf xe = t r + fm ð1Þ Z~c (s)Ge (s)
+ x_ tr
1 + Zf (s)Ge (s) + Z~c (s)Ge (s)
where tr and fm denote the torque of a robot actuator
and the force exerted by a human operator, respec- where Ge (s) denotes the transfer function of the
tively. xr and xe denote the displacement of the posi- distal fractured bone with the muscle, that is,
tioning robot and the displacement of the distal bone, Ge (s) = 1=(me s + be + ke =s).
respectively. mr and br denote the mass and viscosity Figure 4 illustrates the proposed HRRCC scheme,
coefficients of the robot, respectively. bf and kf denote this includes the block diagrams described in Figure 3
the viscosity coefficient and the spring constant of the and additional blocks for the robot controllers and the
distal bone fixation jig, respectively. dynamics of an operator. The basic concept of the
Secondly, the dynamics of the traction device and HRRCC is as follows: x_ m is calculated by multiplying
the traction cable can be described as follows: an admittance gain from Am to fm that is measured by
an operational F/T sensor. Similarly, x_ s is calculated by
mtr x€tr + (btr + bc )x_ tr + kc xtr  bc x_ e  kc xe = t tr ð2Þ
multiplying an admittance gain from As to fs, which is
measured by an interaction F/T sensor. The desired
where t tr denotes the torque of a traction device actua- velocity of the E–E x_ r_d is defined by the subtraction of
tor. xtr denotes the displacement of the traction device. x_ s from x_ m, that is, x_ r_d = x_ m – x_ s. The robot’s E–E is
mtr and btr denote the mass and viscosity coefficients of adjusted to follow x_ r_d at each sampling time. This con-
the traction device, respectively. bc and kc denote the trol architecture allows the scale of fm to fs to be adjus-
viscosity coefficient and the spring constant of the trac- table in proportion to the ratio of As to Am.10 These
tion cable, respectively. blocks are marked by a red dashed line in Figure 4 and
Kim et al. 701

Figure 4. Block diagram of a proposed human-robot-robot cooperative control scheme: (a) blocks for positioning robot and distal
bone fixation jig, (b) blocks for traction device and traction cable, and (c) blocks for fractured bone with muscles, distal bone fixation
jig, and traction cable.

are identical to those described in the group’s former positioning robot’s E–E x_ r can be represented as a
work.10 In addition, by using fs as the input of the trac- function of the operational force fm and the interaction
tion device, local feedback control can be implemented. force fs as follows:
By multiplying the admittance gain Atr to fs, the desired
velocity of the traction device’s E–E x_ tr_d is generated, Cr (s)Gr (s)ðAm (s)fm  As (s)fs Þ
x_ r =
this becomes the input of the traction device controller. 1 + Cr (s)Gr (s)
ð7Þ
Herein, lm and ls denote the degree of direct influence lm Gr (s)fm  ls Gr (s)fs
+
from human forces and environmental forces on the 1 + Cr (s)Gr (s)
actual force applied by the robot actuator, respectively,
these range from 0 to 1. If the robot is non-back-
drivable (e.g. with an industrial robot), lm and ls tend In addition, x_ tr can be represented as follows:
to be closer to 0. Otherwise (e.g. with an impedance-
Ctr (s)Gtr (s) ltr Z~c (s)Gtr (s)
type haptic device), lm and ls tend to be closer to 1 .27 x_ tr = Atr (s)fs  x_ e ð8Þ
In Figure 2, as it is assumed here that the robots are 1 + Ctr (s)Gtr (s) 1 + Ctr (s)Gtr (s)
purely back-drivable, there are no direct influence ratio
blocks like those from Figure 3, that is, lm and ls are Herein, fs can be defined as:
considered to be unity; however, for practicality, the
blocks for lm and ls were added, as the robots are not fs = Zf (s)ðx_ e  x_ r Þ ð9Þ
purely back-drivable. Instead, the positioning robot and
the traction device are modeled with large gear ratios,
from the blocks in group (a) in Figure 4.
so they can be considered to have direct influence ratios
Substituting fs = Zf (s)ðx_ e  x_ r Þ into (8) yields:
close to zero. Following this reasoning, the direct influ-
ence ratio lm was adopted as shown in Figure 4. Ctr (s)Gtr (s)
x_ tr = Atr (s)Zf (s)ðx_ r  x_ e Þ
1 + Ctr (s)Gtr (s)
ð10Þ
Characteristics analysis using simplified model ltr Z~c (s)Gtr (s)
 x_ e
In this subsection, some of the characteristics of the 1 + Ctr (s)Gtr (s)
proposed HRRCC are analyzed.
This is because, for simplicity, the robots are
Ability to minimize load on a positioning robot. First, con- assumed to have large gear ratios and the direct influ-
sider the situation in which the robot and traction ence ratio can then be considered to be zero, that is,
device pull on the distal bone. According to the  = Ctr (s)Gtr (s) ,
ltr = 0. In addition, we represent G(s) 1 + Ctr (s)Gtr (s)
whole block diagram in Figure 4, the velocity of the and then (10) can be represented as:
702 Proc IMechE Part H: J Engineering in Medicine 236(5)

x_ tr = Gtr Atr (s)Zf (s)ðx_ r  x_ e Þ ð11Þ cable dynamics no longer affect the system’s dynamics,
this means we can assume that the spring and damping
coefficients of the cable are zero, that is, Z~c (s)’0. In
From the blocks in group (c) in Figure 4, x_ e can be rep-
this case, (12) can be rewritten as:
resented as:

x_ e = Ge (s)Zf (s)ðx_ r  x_ e Þ ð16Þ
x_ e = Ge (s) Zf (s)ðx_ r  x_ e Þ + Z~c (s)ðx_ tr  x_ e Þ ð12Þ

Assuming the robots are not backdrivable, (7) can be

By substituting (11) into (12), (12) can be represented represented as follows:
as:
ðAm (s)fm  As (s)fs ÞCr (s)Gr (s)
x_ e = Ge (s)Zf (s)ðx_ r  x_ e Þ + Ge (s)Z~c (s) x_ r = ð17Þ

ð13Þ 1 + Cr (s)Gr (s)
Gtr Atr (s)Zf (s)ðx_ r  x_ e Þ  x_ e
By substituting (17) into (16), (16) can be rewritten as:
(13) can then be derived as:
ðAm (s)fm  As (s)fs ÞCr (s)Gr (s)Zf (s)Ge (s)
x_ e =
Zf (s)Ge (s) + Zf (s)Atr (s)Gtr (s)Z~c (s)Ge (s) 1 + Cr (s)Gr (s) + Zf (s)Ge (s) + Cr (s)Gr (s)Zf (s)Ge (s)
x_ e = x_ r
1 + Zf (s)Ge (s) + Z~c (s)Ge (s) + Zf (s)Atr (s)Gtr (s)Z~c (s)Ge (s) ð18Þ
ð14Þ
By substituting (9) into (16), (19) can be obtained:
By substituting x_ e from (14) into (9), fs can be repre-
sented as: x_ e = Ge (s)fs ð19Þ

Zf (s) + Zf (s)Z~c (s)Ge (s)

fs = x_ r By substituting (19) into (18), (18) can be rewritten as:
1 + Zf (s)Ge (s) + Z~c (s)Ge (s) + Zf (s)Atr (s)Gtr (s)Z~c (s)Ge (s)
ð15Þ ðAm (s)fm  As (s)fs ÞCr (s)Gr (s)Zf (s)
fs =
1 + Cr (s)Gr (s) + Zf (s)Ge (s) + Cr (s)Gr (s)Zf (s)Ge (s)

If the control gains of the traction device, Ctr(s), are ð20Þ

chosen properly, and at least at a low frequency, the
By rewriting (20) for fs/fm, (21) can be obtained as:
fs Cr (s)Gr (s)Zf (s)Am (s)
=
fm 1 + Cr (s)Gr (s) + Zf (s)Ge (s) + Cr (s)Gr (s)Zf (s)Ge (s) + Cr (s)Gr (s)Zf (s)As (s)
ð21Þ
traction device follows the input signal sufficiently well,
then we can say Gtr (s)’1 to achieve further simplifica- If the control gains of the positioning robot Cr(s)
tion. Then, we find that the transfer functions from x_ r are chosen properly, at least at low frequencies,
to fs and xr to fs become, respectively, a type 0 system the robot will follow the input signal well, resulting
where no integrator exits and a type 1 system where in 1 \ \ Cr(s)Gr(s), this in turn means that
one integrator is included.28 This means that, when Cr (s)Gr (s) 
1 + Cr (s)Gr (s) = Gr (s)’1. The feedback block consisting
there is motion of the robot’s E–E, an interaction force
of Cr (s) and Gr (s) can be replaced by Gr (s) in Figure 4
will exist. However, the positioning robot’s E–E will
move slowly, then fs becomes small, that is, its steady and Gr (s) can be considered as a position-controlled
state is zero. robot transfer function in the general case. When two
Based on this result, it is expected that the load on fractured bones make contact, the stiffness and viscous
the positioning robot is reduced to zero and that all of coefficients of Ge(s) increase to those of the proximal
the muscle forces are managed by the traction device, bone. Therefore, at a low frequency, we can assume
this means that ftr becomes equivalent to fe. that |Ge(s)| \ \ 1. Finally, if a fixation jig with high
impedance is used, that is, Zf(s) .. 1, we obtain (22),
which is similar to the result described in Ye and
Ability to scale up/down the force feedback. Next, consider a Chen10:
situation in which the robot and traction device release
fs Am (s)
the distal bone, resulting in the muscle becoming ffi ð22Þ
fm As (s)
relaxed (muscular contraction release). When the mus-
cle is relaxed and the two fractured bones are aligned
well, the fractured bones come into contact. In this situ- Therefore, for a RAFR system with the proposed
ation, the cable tension becomes zero, and then the HRRCC, while pulling a distal fractured bone
Kim et al. 703

and silicone display some of the same elastic physical

characteristics, it is difficult to accurately mimic
the behavior of real muscles around the femur using
silicone materials because their physical behavior is dif-
ferent. However, the purpose of this experiment is not
to accurately simulate the contractile force of the
muscles on the femur, it is instead to observe whether
the traction robot assists by minimizing the axial load
on the bones by providing appropriate traction in real-
time to cancel the axial load on the positioning robot.
Therefore, the external contractile force necessary for
this experiment was generated using silicone material.
The RAFR system consisted of a six-axis serial
type positioning robot developed by Hyundai Heavy
Industries Co., Ltd. (HHI), a single-axis motorized
traction device developed by Taekyung Automation
Figure 5. Overview of the experimental setup. A robot- System Co. and Chonnam National University, and a
assisted long bone fracture reduction system consisting of a
3-D image medical navigation system developed by
6-DOF positioning robot, a 1-DOF traction device, a fractured
Corelinesoft Co., Ltd.
bone with muscles, and a 3-D image navigation system:
(1) Fractured femur bone with phantom muscles, (2) 6-DOF
positioning robot, (3) 1-DOF traction device, (4) Traction cable,
1-DOF traction device
(5) 3-D image navigation s/w, (6) Optical tracking system,
(7) Operational F/T sensor, (8) Interaction F/T sensor, The 1-DOF traction device was attached to a support-
(9) Reference maker for distal bone, (10) Reference marker for ing fixture on the side of the positioning robot. The
proximal bone, (11) Proximal bone fixation jig, (12) Distal bone position of the traction device was adjusted to align the
fixation jig, (13) Distal medical pins, (14) Proximal medical pins, axial axis of the proximal bone and the traction direc-
and (15) Bespoken Steinman pin holder and Steinmann pins. tion following the procedure described in Mukherjee
et al.9 As explained, the traction device was attached to
the distal fractured bone in parallel to the positioning
surrounded by muscles, the traction device is controlled robot. The connection procedure was as follows. First,
with a view to minimizing the axial load on the posi- a Steinmann pin was used to pierce through the most
tioning robot. On the other hand, if two fractured distal part of the distal fractured bone, both sides of the
bones collide, the cable’s tension becomes zero, this pin were inserted into a bespoke pin holder. The pin
results in the positioning robot providing feedback on holder was connected to the E-E of the traction device
the contact force with adjustable scaling fs/fm = Am(s)/ through a cable. As the OA Foundation suggested in,29
As(s). the proximal fractured bone’s axial axis was aligned
with the tractional direction. The tractional direction
Experimental setup was set in the opposite direction of the muscular con-
traction to oppose those forces. As shown in Figure 6,
In this section, we describe an experimental setup to the 1-DOF traction device consists of four parts: an
verify the performance of the proposed HRRCC actuator, a traction part, a controller, and a supporting
scheme. For these experiments, the RAFR system cur- fixture. The traction part was made of a single-axis lin-
rently being developed by the research consortium ear motion guide with a ball screw platform (Misumi
mentioned above, was utilized.3,14 Co., pitch: 10 mm, length: 350 mm, maximum traction
force: 300 N). A servo motor (Maxon Motor Co., Ltd,
RE 60W DC motor, gear ratio 21:1) was selected as the
Overview of the experimental setup actuator. As shown in Figure 6(a), an EPOS2 from
Figure 5 shows an overview of the experimental setup. Maxon Motor Co. was used as the motor controller, it
The femur model (Pacific Research Laboratories, Inc., was connected to the main controller PC by a USB
composite femur, left, medium size) was prepared by cable. A load cell was mounted at the end of the trac-
creating a simple fracture in the middle of the femur. tion rod to monitor the traction force exerted. An eye
The phantom muscles between the fractured bones are bolt attached at the load cell was connected to the pin
made of a silicon material. The silicon-based phantom holder on the distal fractured bone using a cable. The
muscles were bonded at various locations around the maximum speed of the traction device was set to
fractured bones. By changing the lengths and thick- 419.52 rpm (corresponding to 69.92 mm/s), the accel-
nesses of the phantom muscles, the contraction force eration and deceleration were both set to 100,000 rpm/s.
was set to roughly 15 kN/m. Although both muscles The supporting fixture was designed to be fixed to the
704 Proc IMechE Part H: J Engineering in Medicine 236(5)

(a)

(b)

Figure 6. Hardware configuration of 1-DOF traction device; (a) mechanical structure: 1. Actuator, 2. Traction part, 3. Controller,
4. Supporting fixture (b) communication channel setup between the HRRCC and traction device’s controller.

supporting post on the side of the positioning robot’s was used to measure the force exerted on the position-
base. ing robot by the human operator. An interaction F/T
sensor (Delta SI-660-60) from ATI Co., Ltd was also
used to measure the force exerted on the positioning
Fixation of fractured bones robot by the environment.
To fix the fractured bones to fixation jigs, several medi-
cal pins were used. First, the pins were fixed to each
part of the fractured bone’s shaft, and then the other Setup of dynamic reference markers
sides of the pins were fixed to C-shaped bone fixation To utilize the 3-D image navigation system, dynamic
jigs. The jigs were fixed onto the frame of the patient reference markers were attached to the proximal and
bed or onto the E-E of the positioning robot.13 The distal fractured bones, as shown in Figure 5. During the
pins used were conventional Shantz pins that had been reduction procedure, the positions of the reference mar-
modified to increase their rigidity. The pins were kers were measured by a Polaris Spectra optical tracker
designed and manufactured by Solco Medical Co., system (OTS) from NDI Co., and then the positions
Ltd., who are part of our consortium.30–32 As described and orientations of the fractured bones were estimated
in Kyungpook National University Industry-Academic using the registration results. This locational informa-
Cooperation Foundation,32 the fixation pin is designed tion on the bones was then shown in real-time by the
to be fixed on the one side of the bone in order not to 3-D image navigation system developed by Corelinesoft
interfere a guide wire or an intramedullary rod inside Co., Ltd.
the medullary cavity. To attach the proximal bone fixa-
tion jig to the patient bed, a serial linked holder (Noga
Engineering Ltd., MA-61003) was utilized. Registration setup
As a registration algorithm that finds the relationship
between the OTS coordinate system and the CT coordi-
Utilization of two F/T sensors nate system, a point-to-point matching algorithm based
For the force feedback and force scaling of the pro- on feature points was temporarily used. Three K-wires
posed cooperative control, two F/T sensors were with diameters of 1.0 mm were inserted before taking
installed on the last link of the positioning robot, as the CT image. The real positions of the inserted points
described in.10 In this implementation, an operational were measured using a probe from the OTS system, the
F/T sensor (Gamma SI-130-10) from ATI Co., Ltd. positions in the CT were then identified using the 3-D
Kim et al. 705

Table 1. Nomenclature for human-robot-robot cooperative Table 2. Experimental parameters.

control.
Item Specification (Unit)
Symbol Descriptions
Stiffness of phantom muscle 15 (kN/m)
fop Force exerted by human muscles Stiffness of bone fixation Jig 72.9 (kN/m)
fm Force measured by operational force sensor Stiffness of traction cable 129.2 (kN/m)
(master force sensor) Compressive strength of 157 (Mpa)
fs Force measured by interaction force sensor composite bone
(slave force sensor) Compressive modulus of 16.7 (Gpa)
ftr Force exerted by traction device composite bone
Gr Dynamics of positioning robot
Gtr Dynamics of traction device
Ge Dynamics of distal fractured bone with muscles
tr Actuating torque exerted by positioning robot’s stiffness of the bone fixture jig was measured at
controller
72.9 3 103 (N/m),10 while the stiffness of the traction
t tr Actuating torque exerted by traction device’s
controller cable (f 4 mm, steel wire rope) was measured to be
Cr Controller of positioning robot 129.2 3 103 (N/m). All the above-mentioned para-
Ctr Controller of traction device meters, including the stiffness of the silicon muscle, are
Am Admittance gain corresponding to fm in listed in Tables 1 and 2.
positioning robot system
As Admittance gain corresponding to fs in
positioning robot system
Atr Admittance gain corresponding to fs in traction
Experimental results
device system In this chapter, the results from the experimental
xr Displacement of robot’s end-effector
xtr Displacement of traction device’s end-effector evaluation of the proposed HRRCC scheme are
xe Displacement of distal bone presented.
x_ r Velocity of robot’s end-effector
x_ tr Velocity of traction device’s end-effector
x_ e Velocity of distal bone Decrease in axial load on robot
x_ m Velocity converted by multiplying an admittance
gain Am to fm The purpose of this experiment was to show that the
x_ s Velocity converted by multiplying an admittance proposed HRRCC scheme can decrease the axial load
gain As to fs on the positioning robot while a surgeon manipulates
Zh Impedance of human arm (surgeon’s arm) the positioning robot along with the traction device.
Zf Impedance of distal bone fixation jig
Zc Impedance of traction cable
Four subjects participated and were asked to conduct
lm Direct influence ratio of fm to Gr the fracture reduction procedure in two different ways:
ls Direct influence ratio of fs to Gr (Case A) used the positioning robot only and (Case B)
ltr Direct influence ratio of ftr to Gtr used both the positioning robot and the traction device.
In this experiment, the subjects used the 3-D medical
navigation system views but did not observe the bones
directly. In each case, the initial values and final values
image navigation system. The registration errors of the
of the loads on all six axes of the positioning robot were
proximal and distal fractured bones were measured to
measured and used to compare the changes in load.
be less than 0.5 mm.
The force and torque values from the operational F/T
sensor and interaction force F/T sensor as well as the
Selection of experimental parameters force values from the load cell mounted on the traction
device were also recorded for the analysis.
To evaluate the performance of the HRRCC, two frac-
tured bones were placed with certain initial offsets. In
the axial direction, the proximal and distal fracture Comparison of the increase in each axis’ actuator
bones overlapped by about 50 mm, the distal bone was load. Figure 7 shows comparisons of the average
rotated by about 15° along with the AP, lateral, and increase in individual actuator loads on the positioning
axial axes with respect to the proximal fractured bone. robot from before to after performing the fracture
The fracture reduction procedure was carried out until reduction process over the two cases. Please note that
the translation and rotation errors reached less than the positioning robot in its current setup provides per-
2.0mm and 2.0°, respectively. Force feedback was acti- centage values of the actuator loads with respect to its
vated along three translational axes, but torque feed- maximum allowed power. In addition, the controller
back was not utilized in this experiment. The force system of the positioning robot automatically ceases
scale of the force feedback was set as one-to-one by set- providing actuation power when the actuator load
ting both the translational admittance gains of Am and reaches 100%. As such, in this analysis, comparison
As to 0.45 3 1023 m/Ns and setting both the rotational using percentage values is useful to understand when
admittance gains of Am and As to 0.35 rad/Nms. The critical situations occur. In Case A(using the
706 Proc IMechE Part H: J Engineering in Medicine 236(5)

axis’ load increased by only 2.68%, a reduction of 97%

compared to Case A.

Comparison of measured forces from the Two F/T sensors and

load cell. In this subsection, the loads experienced by
each robot will be discussed. Figure 8(a) and (b) show
the measured forces during the traction task of the frac-
ture reduction procedure in Case A and Case B, respec-
tively. The operational force Fm is represented as a red
dashed line, the interaction force Fs as a blue solid line,
and the traction force Ftr as a green dotted line. While
the distal fractured bone was pulled by the positioning
robot (during the traction task), Fm and Fs were mea-
sured, the results are plotted in Figure 8(a). During
human operation of the positioning robot, when the
distal fractured bone was being pulled, the Fs signal is
positive due to the contraction force of the surrounding
Figure 7. Comparisons of the increases in the six axes’
muscles. Conversely, if the positioning robot moves the
actuation loads on positioning robot after fracture reduction in
two cases: (Case A) using positioning robot only and (Case B) bone in the opposite direction bringing the bones into
using both the positioning robot and traction device. In Case A, contact, the signal becomes negative. The maximum
the average increase of all six axes’ actuation loads measured force measured was 158.8 N. In Case B, as shown by
38.1%. The fifth actuation load increased by 91.14% from the Figure 8(b), Fs maintained an average of 0.6 N, whereas
initial value to after the reduction procedure was complete. In Ftr was measured up to a maximum of 172.6 N. As the
Case B (using both the positioning robot and traction device), 1-DOF traction device opposes most of the muscular
the average increase measured only 1.4%. The fifth axis’s load contraction force along the longitudinal axis of the
increased by only 2.68%, a 97% reduction compared to Case A. bone, the load on the positioning robot is reduced sig-
nificantly. The maximum interaction forces experienced
in the individual experiments are listed in Table 3 for
positioning robot only), the average increase for all six comparison.
axes’ actuation loads was measured to be 38.1%. The Figure 9 shows the trajectory profile of the position-
fifth actuation load increased by 91.14% from the ini- ing robot and the traction device as the fracture reduc-
tial value after the reduction procedure. In Case B tion operation progresses from the initial positioning of
(using both the positioning robot and traction device), the fractured bone to completion of fracture alignment
the average increase measured was only 1.4%. The fifth according to the human’s operational force-torque

Figure 8. Comparisons of forces during the traction task part of the fracture reduction procedure: the operational force Fm, the
interaction force Fs, and the traction force Ftr are shown in (a) Case A – using only the positioning robot and (b) Case B – using both
the positioning robot and traction device: Fs measured up to a maximum of 158.8 N while conducting the reduction task in (a). Fs
averaged 0.6 N, whereas Ftr measured up to a maximum of 172.6 N in (b). As the 1-DOF traction device manages most of the
muscular contraction force along the longitudinal axis of the bone, the load on the positioning robot is reduced significantly
eliminating the chance of it becoming overloaded.
Kim et al. 707

Table 3. Comparison of the Measured Maximum Interaction Forces (Fs) during the Traction Task in Fracture Reduction Procedure
for Case A and Case B.

Subject 1 2 3 4 Average 6 STDEV

Max. Inter. Forces Fs (N)

Case A (Positioning Robot Only) 154.7 158.8 180.0 161.0 163.63 6 9.7
Case B (Positioning Robot + Traction Device) 14.32 11.39 15.94 15.19 14.21 6 2.0

and also to determine the level of transparency of the

force feedback. To demonstrate the force reduction
effectiveness, experiments were performed for two
cases: Case A, using the positioning robot only without
a force feedback function, and Case B, using both the
positioning robot and the traction device while the pro-
posed HRRCC scheme was active. In the experiment,
for simplicity, the motion of the positioning robot was
constrained to 1-DOF motion along the longitudinal
axis of the bone, the 1-DOF interaction force was mea-
sured in both cases. For each case, the operational
force Fm, the interaction force Fs, and the displacement
of the positioning robot were measured. Figure 10
shows the force and displacement variations after the
two fractured bones made contact. Please note that the
data in Figure 10 were recorded in the situation where
no visual feedback was provided, this was done to
Figure 9. Reduction trajectory profile: (a)–(c) the translational clearly demonstrate the effect of HRRCC. Fm is repre-
positions (x, y, z) of the broken end of the distal fractured bone sented by a red dashed line, and Fs is represented by a
with respect to the positioning robot’s base frame (d) the blue solid line. Figure 10(a) shows the forces in Case A
position of the end of the traction device with respect to the (without force feedback). Since the fractured bones
traction device’s base frame, (e) the x-axis direction’s interaction were separated initially, Fm could be recorded from
force Fs. (f) the traction force Ftr. t = 1 s while Fs remained zero. Note that since admit-
tance control without a force feedback function was
utilized, the sign of Fm is related to the velocity of the
commands. (a) to (c) Show the translational positions robot’s E–E, and thus it is roughly proportional to the
(x, y, z) of the distal fractured bone’s broken side with variation of the interaction force fs. When contact with
respect to the position of the robot base frame. It the bone was released, Fs became zero again, and only
is mapped as the positioning robot’s tool position. Fm could be measured. In this experiment, an Fs of up
(d) Shows the position of the end-effector of the trac- to 2220 N could be recorded because no feedback
tion device with respect to the traction base frame, and information was provided, including by the 3-D image
(e) shows the x-axis direction’s interaction force Fs, and navigation system. The value of 2220N is almost the
(f) shows the traction force Ftr. This graph also shows maximum force against which the temporary bone fixa-
that the axis’ actuator load on the position robot’s end- tion jig that was used can resist. If a stronger fixation
effector is reduced thanks to the traction force assis- jig is used, it is expected that this maximum value could
tance. The maximum traction force was measured to be be higher. Due to the considerable power of the robot,
146.7 N. From the time when the maximum traction the bones and mechanical structure cannot maintain
force was measured to the completion of fractured their positions or orientations. Thus, there was a large
bone alignment, the total displacement of the bone held displacement of the robot’s position, as shown in graph
by the positioning device was measured to be 10.1 mm Figure 10(c).
and the displacement of the traction device was mea- Figure 10(b) shows the results from Case B (with
sured to be 10.3 mm. force feedback using the proposed HRRCC). Even
after the bones are placed in contact, the positioning
robot continued to exert force on the broken bone. The
Force feedback function of HRRCC interaction force Fs caused the robot to move back-
The purpose of this experiment was to verify whether ward, which reduced the excess force acting on it.
the force feedback function of the proposed HRRCC Thus, the interaction force Fs maintained a similar
scheme reduces the excessive force required of the posi- value to the operational force input Fm. Since Fs acts as
tioning robot during the fracture reduction procedure a resistive force against the robot’s motion, it was also
708 Proc IMechE Part H: J Engineering in Medicine 236(5)

(a) (b)

(c) (d)

Figure 10. Force feedback evaluation while fractured bones have been placed in contact during the fracture reduction procedure:
(a) force comparison of Fm and Fs and (c) position profile of positioning robot in Case A, (b) force comparison and (d) position profile
of the positioning robot in Case B (with 1-DOF force feedback). Without force feedback, an operational force of about 22.6 N could
move the bone up to 33 mm, resulting in the largest interaction force reaching 2227 N. With force feedback, an operational force of
about 234.02 N could move the bone up to 3.99 mm, resulting in the largest interaction force reaching –34.13 N. The force feedback
control prevented the robot from applying the strong forces to the proximal bone seen in (c). Note that this data was recorded in a
situation where no visual feedback was provided to more clearly understand the effects of force feedback.

Table 4. Comparison of the measured maximum interaction forces (Fs) while fractured bones placed into contact during fracture
reduction procedure for Case A and Case B. Note that 3-D Image navigation views were provided in these experiments.

Subjects 1 2 3 4 Average 6 STDEV

Case A
Max. Fs (N) 225.03 219.57 230.7 217.88 223.3 6 5.8
Max. position profile (mm) 3.32 2.54 4.1 2.34 3.8 6 0.8
Case B
Max. Fs (N) 217.45 27.24 212.71 216.31 213.4 6 4.6
Max. position profile (mm) 2.13 1.17 1.66 2.71 1.92 6 0.7

observed that the operators tended to naturally reduce subject carefully moved the robot; thus, the maximum
the forces they applied. force was decreased compared to earlier experiments.
Based on this preliminary test, four participating Although the interaction force was reduced after pro-
subjects were then asked to conduct the task of placing viding visual feedback, the maximum interaction force
the bones in contact from the fracture reduction proce- decreased substantially when force feedback was also
dure by directly manipulating the positioning robot. provided. The X-axis displacement of the proximal
Table 4 lists the maximum interaction forces Fs as well fractured bone in Case A measured 3.8 6 0.8 mm, while
as the maximum X-axis’ displacement of the position- in Case B it measured 1.92 6 0.7 mm, a 49% decrease
ing robot for two cases: Case A, without 1-DOF force in force.
feedback from the HRRCC scheme but with visual
feedback, and Case B, with both force feedback and
Conclusion
visual feedback. The interaction forces in Case A mea-
sured 223.3 6 5.8 N, while those in Case B measured Our research consortium has developed a robot-assisted
13.4 6 4.6 N. As visual feedback was provided, the long bone reduction system, which consists of a serial
Kim et al. 709

type manipulator in the form of a positioning robot to publication of this article: This work was supported by
accurately position the fractured bones during the the National Strategic R&D Program for Industrial
reduction procedure while a traction device is used Technology (No. 10041605) funded by the Ministry of
simultaneously to combat the considerable contraction Trade, Industry and Energy (MOTIE), South Korea
forces from the surrounding muscles that would other- and also by the Korea Medical Device Development
wise make such procedures much more difficult. To Fund grant funded by the Korean government (the
effectively control a 6-DOF serial type positioning Ministry of Science and ICT, the Ministry of Trade,
robot and a 1-DOF traction device through a surgeon’s Industry and Energy, the Ministry of Health and
operational command inputs, a human-robot-robot Welfare, the Ministry of Food and Drug Safety) (proj-
cooperative control (HRRCC) scheme is proposed. In ect number: KMDF_PR_2020901_0122, 1711138226).
the proposed HRRCC scheme, the robots are physi-
cally connected, and they share force information ORCID iDs
recorded by sensors for feedback control. To measure
the operational and interaction forces separately, two Woo Young Kim https://orcid.org/0000-0001-8355-
F/T sensors were installed on the end-effector of the 5113
positioning robot. By applying admittance control to Seong Young Ko https://orcid.org/0000-0003-4316-
the two forces, the positioning robot can be manipu- 0074
lated directly, while a force feedback function is imple-
mented for bone contact situations. Furthermore, a References
1-DOF motorized traction device is actuated by apply- 1. Gösling T, Westphal R, Faülstich J, et al. Forces and
ing another manner of admittance control as feedback torques during fracture reduction: intraoperative
from the interaction forces. The use of the developed measurements in the femur. J Orthop Res 2006; 24:
robot-assisted fracture reduction system showed that 333–338.
the 1-DOF traction device is able to manage most of 2. Joung S, Park C-W, Park J, et al. Strategy for robotiza-
the muscular contraction forces experienced while posi- tion of lower limb fracture reduction. In: The 44th inter-
tioning a fractured bone, this allows the positioning national symposium on Robotics (ISR), Seoul, Korea,
robot to conduct the reduction task without fighting October 24–26, 2013.
large axial loads. When the fracture bones are placed in 3. Lee S-H, Jung S-H, Lee C-H, et al. Development of a
Surgical robotic system for repairing damaged musculos-
contact, the positioning robot acts as though is no trac-
keletal region. In: The 44th international symposium on
tion device, allowing it to provide scaled force feedback robotics (ISR), Seoul, Korea, 24–26 October 2013.
in consideration of the operational and interaction forces. 4. Cho M, Jung S, Lim H, et al. Development of a 6-axis
Performance in these tasks was verified using 1-DOF articulated robot for the fracture reduction surgery. In:
mathematical modeling and a working prototype. The 44th international symposium on robotics (ISR),
In future work, we need to expand our theoretical Seoul, Korea, 24–26 October 2013.
modeling to formulate a general case describing an n- 5. Westphal R, Winkelbach S, Wahl F, et al. Robot-assisted
DOF positioning robot and traction device. Theoretical long bone fracture reduction. Int J Rob Res 2009; 28(10):
analyses regarding the range of the admittance gains 1259–1278.
and the stability and transparency of the proposed 6. Joung S, Liao H, Kobayashi E, et al. Hazard analysis of
HRRCC system should also be carried out. In addition, fracture-reduction robot and its application to safety
design of fracture-reduction assisting robotic system. In:
animal tests are expected to be conducted using a work-
2010 IEEE international conference on robotics and auto-
ing prototype.
mation, Anchorage, AK, USA, 3–8 May 2010. New
York: IEEE.
Acknowledgements 7. Mukherjee S, Rendsburg M and Xu WL. Surgeon
instructed, image-guided and robot-assisted long bone
The authors thank the research consortium, especially
fractures reduction. In: 1st International conference on
Hyundai Heavy Industry Co., Solco Biomedical Co., sensing technology, Palmerston North, New Zealand,
Ltd., Taekyung Automation System Co., and 2005.
Corelinesoft Co., Ltd., for their support of this work. 8. Ye R and Chen Y. Development of a six degree of free-
dom (DOF) hybrid robot for femur shaft fracture reduc-
Declaration of conflicting interests tion. In: Proceedings of the IEEE international conference
on robotics and biomimetics, Bangkok, Thailand, 21–26
The author(s) declared no potential conflicts of interest February 2009. New York: IEEE
with respect to the research, authorship, and/or publi- 9. Dagnino G, Georgilas I, Tarassoli P, et al. Vision-based
cation of this article. real-time position control of a semi-automated system
for robot-assisted joint fracture surgery. Int J Comput
Assist Radiol Surg 2016; 11(3): 437–455.
Funding
10. Kim WY, Ko SY, Park J-O, et al. 6-DOF force feedback
The author(s) disclosed receipt of the following finan- control of robot-assisted bone fracture reduction system
cial support for the research, authorship, and/or using double F/T sensors and adjustable admittances to
710 Proc IMechE Part H: J Engineering in Medicine 236(5)

protect bones against damage. Mechatronics 2016; 35: 21. Cui R and Yan W. Mutual synchronization of multiple
136–147. robot manipulators with unknown dynamics. J Intell
11. Bowyer SA, Davies and BL, Rodriguez y Baena F. Robot Syst 2012; 68(2): 105–119.
Active constraints/virtual fixtures: a survey. IEEE Trans 22. Sun D. Position synchronization of multiple motion axes with
Robot 2014; 30: 138–157. adaptive coupling control. Automatica 2003; 39(6): 997–1005.
12. He X, Balicki M, Gehlbach P, et al. A multi-function 23. Liu H. H T and Sun D. Uniform synchronization in multi-
force sensing instrument for variable admittance robot axis motion control uniform synchronization in multi-axis
control in retinal microsurgery. In: 2014 IEEE interna- motion control. Proc Amer Control Conf 2005; 7: 4537–4542.
tional conference on robotics & automation (ICRA), Hong 24. Chen CS and Chen LY. Cross-coupling position com-
Kong, China, 31 May–7 June 2014. New York: IEEE. mand shaping control in a multi-axis motion system.
13. Park IH, Oh CW, Joung S. Bone traction apparatus and Mechatronics 2011; 21(3): 625–632.
fracture reduction apparatus having the same. Patent 25. Li Y and Ge SS. Force tracking control for motion syn-
101564717, Korea, 2015. chronization in human-robot collaboration. Robotica
14. Joung S, Park CW, Kim KR, et al. Development of a 2016; 34(6): 1260–1281.
bone positioning robotic system for fracture reduction. 26. Liu C, Zhao D and Xiang X. Force synchronization of
In: 29th ICROS Annual Conference, Daegu, Korea, 29 multiple robot manipulators with uncertain contact stiff-
May–30 2014. ness. In: Proceedings of the 34th Chinese control confer-
15. Zhu Q, Sun XG, Wang XS, et al. Development of intrao- ence, Hangzhou, China. pp. 1173–1178, 2015.
perative noninvasive force measuring system during 27. Abbott JJ and Okamura AM. Pseudo-admittance bilat-
femoral fracture reduction. In Proceedings of 10th annual eral telemanipulation with guidance virtual fixtures. In:
IEEE international symposium on medical measurements Proceedings of the haptic interfaces for virtual environment
and applications, Turin, Italy, 2015, pp. 335–339. New and teleoperator systems symposium, 25–26 March, Alex-
York: IEEE. andria, VA, pp. 169–175, 2006.
16. Georgilas I, Dagnino G, Tarassoli P, et al. Preliminary 28. Ogata K. Modern Control Engineering. 5th ed. Upper
analysis of Force–Torque measurements for robot- Saddle River, NJ: Prentice Hall, 2010.
assisted fracture surgery. In: Annual conference of the 29. Banerjee R, Brink P, Cimerman M, Pohlemann T, Toma-
IEEE engineering in medicine and biology society (EMBC zevic M. Skeleton tractionReduction and fixation, AO
2015), Milan, Italy, 2015. New York: IEEE. Foundation, https://surgeryreference.aofoundation.org/
17. Tarn TJ, Bejczy AK and Yun X. Coordinated control of orthopedic-trauma/adult-trauma/pelvic-ring/llium/extern
two arms. In: Proceedings of IEEE international confer- al-fixation-and-traction (2018, accessed 1 February 2022)
ence on robotics and automation, Arlington, VA, 1986, pp. 30. Park IH, Oh CW, Lee HJ, et al. Fixation pin for frac-
1193–1202. New York: IEEE. ture-reduction with intramedullary nailing system. Patent
18. Hayati S. Hybrid Position/Force Control of Multi-Arm 101652286, Korea, 2016.
Cooperating Robots. In: Proceedings of the 1986 IEEE 31. Kyungpook National University Industry-Academic
international conference on robotics and automation, pp. Cooperation Foundation. Bone traction apparatus and
82-89, 1986. New York: IEEE. fracture reduction apparatus having the same. Korea pat-
19. Sun D and Mills JK. Adaptive synchronized control for ent no. 1015647170000, 2015. DOI: 10.8080/10201400
coordination of multi-robot assembly tasks. IEEE Trans 40001?urlappend=en
Robot Autom 2002; 18(4): 498–510. 32. Kyungpook National University Industry-Academic
20. Zhao D, Li S and Zhu Q. Adaptive Jacobian synchro- Cooperation Foundation. Fixation Pin for fracture-
nized tracking control for multiple robotic manipulators. reduction with intramedullary nailing system. Korea pat-
In: Proceedings of Chinese Control Conference, Yantai, ent no. 1016522860000, 2016. DOI: 10.8080/102014007
China pp. 3705–3710, 2011. 2246?urlappend=en