Content uploaded by Liao Wu
Author content
All content in this area was uploaded by Liao Wu on Aug 15, 2021
Content may be subject to copyright.
SnakeRaven: Teleoperation of a 3D Printed Snake-like Manipulator
Integrated to the RAVEN II Surgical Robot
Andrew Razjigaev, Ajay K. Pandey, David Howard, Jonathan Roberts Senior Member IEEE,
and Liao Wu Member IEEE
Abstract— Telerobotic systems combined with miniaturised
snake-like or elephant-trunk robotic arms can improve the
ergonomics and accessibility in minimally invasive surgical
tasks such as knee arthroscopy. Such systems, however, are
usually designed in a specific and integral approach, making it
expensive to adapt to various procedures or patient anatomies.
3D printed instruments with a detachable design can bring
the benefits of patient-specific customisation, affordability, and
adaptability to new clinical scenarios. However, the integration
of such snake-like instruments to standard telerobotic systems
can be challenging in terms of design and control. In this study,
a teleoperation system is developed to control and steer the
pose of SnakeRaven: a 3D printed, customisable snake-like end-
effector attached to the RAVEN II platform for the application
of fibre-optic knee arthroscopy. Algorithms for the parametric
inverse kinematics and mapping between the RAVEN II joint
space to the coupled tendon-driven rolling joints are developed.
The controller is tested and validated on the physical prototype
interfacing with the RAVEN II platform in a teleoperation
experiment. A video demonstrating the main results of this
paper can be found via https://youtu.be/ApJjR853kIQ
I. INTRODUCTION
In a recent survey, surgeons believe that knee arthroscopy
is a difficult procedure that is ergonomically challenging with
frustrating tools and technology [1]. A telerobotic platform
could be a solution as it can provide an ergonomic system
with visualisation enhancement in arthroscopy. Surgical teler-
obotic platforms such as the da Vinci and the RAVEN have
already enhanced the technical capabilities for procedures in
Minimally Invasive Surgery (MIS). The RAVEN telerobotic
platform developed by the Biorobotics Laboratory at the
University of Washington was designed to be a modular and
light-weight solution for remote teleoperation [2]. The most
recent version, the RAVEN II, has 7 Degrees Of Freedom
(DOF) with instruments constrained about a Remote Center
of Motion (RCM). It is similar in functionality and structure
to the da Vinci surgical robot and made for the application
of laparoscopy [3]. It is an open-source platform for col-
laborative research which encourages new implementations,
modifications and improvements to expand the capabilities of
telerobotic systems. Current surgical telerobotic platforms in
A. Razjigaev, A. K. Pandey, and J. Roberts are with the School of
Electrical Engineering and Robotics at the Queensland University of Tech-
nology and the Australian Centre for Robotic Vision, Brisbane, Australia.
a.razjigaev@qut.edu.au
D. Howard is with the Cyber Physical Systems Program, Data61,
CSIRO. david.howard@csiro.au
L. Wu (corresponding) is with the School of Mechanical and Manu-
facturing Engineering, University of New South Wales, Sydney, Australia.
dr.liao.wu@ieee.org
Fig. 1. A. This is SnakeRaven - a 3D printed instrument attached to the
RAVEN II telerobotic system. B. a close up of the adaptor to the RAVEN
II and C. a close up of the 3D printed snake-like end-effector
hospitals, however, have a high cost for maintenance and
single-use robotic appliances [4].
Continuum robotic manipulators that resemble biological
system such as elephant trunks, tentacles and snakes, can be
steered to traverse confined spaces, reach difficult-to-access
surgical sites and complete tasks with dexterity [5]. These
snake-like manipulators present an interesting solution for
emerging telerobotic platforms as they can be miniaturised
for flexible access surgery [6]. With additive manufacturing,
they can be made bespoke to a patient allowing greater cus-
tomisation and cost efficiency to the extent that we can afford
to have disposable instruments [7], [8]. It is important that
the field moves towards patient-specific robotics because the
intervention approach can vary significantly across patients
in terms of the workspace volume, access to the intervention
site and the required motion capability [8]. Without tailored
surgical devices, the risk of injury to adjacent organs or
tearing large vessels is greater and therefore it is imperative
to customise the surgical robot to maximise the benefits
of robot-assisted surgery and improve patient outcomes [8].
Therefore integrating 3D printed continuum manipulators
that can easily snap onto to the arms of telerobotic platforms
can allow affordable bespoke interchangeable instruments to
be used for different patient-specific applications.
In this work, we investigate the development of a tele-
operation system for controlling SnakeRaven - our 3D
printed, customisable, and steerable instrument attached to
the RAVEN II platform - for fibre-optic knee arthroscopy.
The results indicate the feasibility of integrating customised
snake-like continuum instruments that can potentially be
optimised for different tasks or patient anatomies. We believe
that this research can further expand the capabilities of the
RAVEN platform or other similar platforms for new surgical
applications and problems in the future.
A. Related Work
Combining a small miniaturised instrument to telerobotic
platforms creates a macro-micro manipulator. The telerobotic
platform acts as a macro-manipulator for workspace cover-
age while the instrument becomes a micro-manipulator for
higher performance. This mechatronic solution can bring
the benefits of telerobotic platforms to arthroscopy such
as ergonomics and more controllability of the DOF in a
continuum robotic arm [9]. Current studies with steerable
tools for arthroscopy are mostly limited to being handheld
with one bending DOF [10]. In the application of hip surgery,
Kutzer et al. developed a steerable cable-driven notched
continuum instrument attached to a robot [11], [12]. With
the da Vinci, Francis et al. developed both a concentric
tube robot and a notched tube robot [13]. Another group
incorporated 3mm pin-joint wristed instruments for da Vinci
which was able to minimise the instrument self-occlusion in
a suturing task [14]. One study developed an organ retractor
with tendon-driven rolling joints actuated by the da Vinci
robot [15]. The only patient-specific robot attachable to da
Vinci is the 3D printable concentric tube robot by Morimoto
et al. [16]. Although concentric tube robots are strong, thin,
stiff and able to manoeuvre through tortuous paths, they lack
dexterity at the tip compared to a tendon-driven mechanism
which is an attribute required in arthroscopy [17].
The only work in the literature about continuum end-
effectors for the RAVEN II is the flexible imaging probe
developed by Carlos et al [18]. Their mechanism uses the de-
formation of a spring to direct an endoscope for visualisation
in telemanipulation tasks [19]. It was later, redesigned to be
attached parallel to the existing robotic grasper instruments
[20] and most recently it was used to visualise an oscillating
phantom [21]. Although they developed a flexible imaging
tool, the simple spring mechanism, shaft diameter and open
loop control law would not transfer to our problem of
teleoperated patient-specific fibre-optic knee arthroscopy.
B. Contribution
The focus of this paper is to determine how to control
SnakeRaven - a patient-specific snake-like manipulator at-
tached to the RAVEN II - for teleoperation tasks that require
control of the position and orientation of the toolpoint. Our
contribution is the development of algorithms for the closed
loop kinematic teleoperation of SnakeRaven given parametric
variation of the design. This closed loop system considers
calculations for the forward kinematics, inverse kinematics
and conversions between the RAVEN II actuation and the
coupled rolling joint actuation. The teleoperation system
is implemented and validated on the RAVEN II platform
through experimentation.
II. PROB LEM FORMULATION
The clinical application of SnakeRaven, in our case study,
is to be used as a tool in laser arthroscopic knee chon-
droplasty [22]. This task involves controlling the position
and orientation of a flexible fibre-optic laser to reach torn
tissue locations inside the knee. Thin and flexible fibres have
a great potential in knee surgery [23], they can efficiently
transmit laser irradiation in aqueous environments and can
simultaneously be used to provide endoscopic imaging [22].
Controlling the tip of the flexible fibre requires a steerable
instrument that can be teleoperated to do the procedure.
A. Control Task Objectives and Assumptions
•The main objective is to have sufficient controllability
of the position and orientation of the tip of the end-
effector from the console interface
•The controller would need to interface with the RAVEN
II and map its motion to the motion of the end-effector
•Given the design parameterisation of the end-effector,
the control algorithm must be able to control any
variation of the end-effector design.
•In this study, it is assumed that the patient-specific
design parameters are given thus the concern is only
the teleoperation aspect of SnakeRaven
•It can be assumed that the fibre-optic cable is light-
weight and highly flexible with negligible stiffness
B. SnakeRaven Macro-Micro Structure Overview
SnakeRaven is a steerable instrument that can be actuated
from a RAVEN II surgical robot.
1) RAVEN II macro arm section properties:
•The RAVEN II steers the instrument about the RCM
with three degrees of freedom (DOF) (shoulder and
elbow rotations with translation)
•An adapter, as seen in Fig. 2 (b) was designed to utilise
all four instrument actuators for the actuation of the
snake arm. Note: since the fibre-optic probe is invariant
to tool roll, a roll actuation is not present
•The adaptor is 3D printed from a Stratasys connex3
objet500 polyjet 3D printer using the Vero brand of
photopolymer.
•The adaptor can apply tension to the tendons by moving
the guide wheels in their slots
Fig. 2. Diagram of the structure of the SnakeRaven a. the two-module design structure and b. interior of the adaptor to the RAVEN II
•To calibrate the setup, the RAVEN II positions the tool
at a pose perpendicular to the table at the RCM and is
inserted top-down into its task-space
2) Snake End-effector micro arm section properties:
The end-effector is actuated by tendon driven rolling joints
because of its advantages of allowing a wide range of
bending, a hollow channel and a shape suitable for small-
scale 3D printing [24], [25]. The model is based on the
rolling joint system seen in [24].
•Due to the size requirements for knee arthroscopy [10],
the end-effector width is 4mm with a 2mm hollow
channel wide enough to accommodate the light-weight
sub-milimetre fibre with negligible influence from the
fibre maximum bending and stiffness
•The overall length of the end-effector is 18mm and its
structure is shown in Fig. 2 (a)
•The tool consists of two bending modules: a distal and
a proximal module whose properties are highlighted in
red and blue respectively in Fig. 2 (a)
•Each module has 2 DOF through bending in the pan or
tilt directions. The two-module end-effector therefore
has 4 DOF and the combined DOF of SnakeRaven is 7
•Each module has three design parameters that can vary
patient-specifically defined in vector p=α n d
where ndefines the number of rolling joints in the
module, dis the distance between each rolling surface
and αis half the angle of curvature for the rolling joint.
•The design parameters of the end-effector in this study
are shown in table I
•To separate the overlap of tendons, the distal module is
offsetted by 45° to the proximal module
•The rolling joints are 3D printed from a Projet 3510SD
by 3D systems with a UV Curable Plastic Visijet M3
Crystal material
TABLE I
END -EFFEC TOR PROP ERTIE S AND PAR AM ET ER S
Design Proximal Distal
Modules DOF α1n1d1α2n2d2
m=2 7 0.20 3 1.00 0.88 3 1.00
III. KIN EM ATI C CONT ROL LOO P FOR SNAK ERAVE N
The SnakeRaven controller was programmed in C++ using
ROS to communicate to the RAVEN II control software [26]
and using a keyboard for mapping position and orientation
directions in the task space. The RAVEN II control software
was modified to have a new joint level control mode where
the joint motion is commanded by joint deltas that come from
a ROS topic. An illustration of this control loop is seen in
Figure 3. The SnakeRaven controller node subscribes to the
RAVEN II node and receives data about the joint angles of
the instrument capstans and publishes an incremental update
computed from the inverse kinematics.
Fig. 3. SnakeRaven control loop interfacing between the RAVEN II and
snake sides of the system (physical reference shown left)
A. Forward Kinematics of SnakeRaven
The forward kinematics for SnakeRaven combines the
three DOF of the RAVEN II about the RCM and the pan
and tilt DOF of the rolling joints. It can be represented as a
function Kthat outputs a transform Tend which depends on
the joint vector qand the parametric design vector p.
Tend =K(q, p)(1)
The forward kinematics is implemented as pseudo-code
in algorithm 1 which is written for a variable amount of
modules. With a two-module design, the total joint vector q
has seven DOF:
q=θ1θ2d3θ1pθ1tθ2pθ2t(2)
The first three θ1, θ2, d3are the joints of the RAVEN II
which are modelled as a set of Denavit Hartenberg (DH)
homogeneous transformations as described in the RAVEN
II kinematic report [27]. Line 2 of algorithm 1 starts the
chain of homogeneous transformations by multiplying the
RCM fixed frame with the DH matrices of the first three
joints. The rest of the DOF in qare bending angles for
the modules of pan and tilt rolling joints. Lines 3 to 17
is a for loop calculating the transformations after each kth
module for mmodules in a system. Lines 4 and 5 extract the
variables from pand qrespectively to compute the rolling
joint transformation matrices Tpan and Ttilt that can be
derived from Fig. 4 in eq. 3 and 4.
Fig. 4. Diagram of the rolling joint model
Tpan =
1 0 0 0
0 cos φp−sin φp−hsin(φp/2) −dsin φp
0 sin φpcos φphcos(φp/2) + dcos φp
0 0 0 1
(3)
Ttilt =
cos φt0 sin φthsin(φt/2) + dsin φt
0 1 0 0
−sin φt0 cos φthcos(φt/2) + dcos φt
0 0 0 1
(4)
Where φp,φtare the individual rolling joint bending angles
computed as the pan or tilt angles divided by the number of
pan or tilt joints θp/np,θt/nt.
A nested for loop from lines 6 to 12 goes through each
ith rolling joint in a module of njoints and multiplies the
transformations to the kinematic chain. The logic statement
in line 7 and the update in line 16, ensure the joints preserve
the alternating pan and tilt pattern. The if statement in lines
13 to 15 applies a rotational offset in between modules and a
fixed transform to the tip of the fibre-optic probe is appended
in line 18.
Algorithm 1 Forward Kinematics of Snake Raven
1: function FORWAR DKIN EM ATI CS(q,p)
2: Tend =RCM ∗T1(q1)∗T2(q2)∗T3(q3)
3: for k=1:mdo
4: α, n, d ←p
5: Tpan, Ttilt ←q4+2(k−1) , q5+2(k−1)
6: for i = FirstJoint : FirstJoint+n do
7: if !(P anF irst XOR isodd(i)) then
8: Tend =Tend ∗Tpan
9: else
10: Tend =Tend ∗Ttilt
11: end if
12: end for
13: if k! = mthen
14: Tend =Tend ∗TRotz(−π/2m)
15: end if
16: Update(P anF irst, F ir stJoint)
17: end for
18: Return =Tend ∗Ttool
19: end function
B. Inverse Kinematic Control of SnakeRaven
The inverse kinematic controller incrementally computes
the joint updates for quntil the forward kinematics Tend
becomes the desired pose Tdes. The incremental delta dx
between Tend and Tdes can be computed as:
dx =tdes −tend
vex(RdesRT
end −I3)(5)
Where tis a translation vector, Ris a rotation matrix,
vex() is the inverse of the skew symmetric matrix operator
and I3is a 3×3identity matrix. This delta can be used to
compute The joint update dq by inverting the manipulator
Jacobian matrix J:
dx =J(q)dq (6)
To calculate the Jacobian parametrically, algorithm 1 was
modified to compute the revolute joint Geometric Jacobians
as cross products between the axis of actuation ωand the
distance to the tip Pgfrom DOF position Piin equation 7:
J(q) =
(ωz×(Pg−Pq1))TωT
z
(ωz×(Pg−Pq2))TωT
z
ωT
z0
(ωx×(Pg−Pθ1p))TωT
x
(ωy×(Pg−Pθ1t))TωT
y
(ωx×(Pg−Pθ2p))TωT
x
(ωy×(Pg−Pθ2t))TωT
y
T
(7)
This matrix Jis 6×7relating each DOF to a translational
and angular velocity. For modules with multiple pan and
tilt joints, the average Jacobian column is computed for
that DOF [24]. This matrix is redundant and is inherently
singular. To invert Jto compute the joint update dq, a
weighted damped least-squares method was implemented:
dq =JT(JJT+λ2I)−1W dx (8)
Where the damping factor of λ= 1 and a weighting matrix
Wis a diagonal matrix of constant weights:
W=diag(111555)(9)
These weights are used to balance the magnitude of
rotation deltas with the translation deltas allowing the solver
to converge to the desired orientation just as fast as it does
to the desired position suitable for our application [28].
C. Joint to Actuation Space
The inverse kinematics solves the joint vector but to send
RAVEN II joint commands, the bending angles in qmust
be solved as capstan rotations jpos. The capstan motion is
defined as being in the actuation space.
To solve the capstan motion, the equations for the change
in tendon displacements in a single pan and tilt module for
the left lland right lrtendons can be derived from Fig. 4 as
equation 10:
∆lpr
l= 2npr(cos(α)−cos(α±θp
2np
)) + 2ntr(1 −cos θt
2nt
)
∆ltr
l= 2ntr(cos(α)−cos(α±θt
2nt
)) + 2npr(1 −cos θp
2np
)
(10)
With capstans that have radius ra, the actuation is there-
fore:
jpos(i) = ∆li(θp(i), θt(i))
ra
(11)
However, this calculation is difficult to implement para-
metrically for many modules. So a solution would be to
measure the tendon length distances directly from the kine-
matic transformations by modifying algorithm 1. Adding
up the tendon lengths and subtracting it from the neutral
configuration can give us the change in tendon displacement
∆lias needed in equation 11.
D. Actuation to Joint Space
The RAVEN II will follow joint commands but when
reading the current joint values from it, the data is in the
actuation space jpos and needs to be mapped back to the
joint space q. Reversing the calculation leads to a series
of coupled variables that makes it difficult to solve. This
requires an iterative approach to solve qwhere an estimate
is incrementally determined until the error converges.
For a simple two DOF module with joint angles θp, θt, the
corresponding motor angles jp, jtfor the left tendons can be
solved with equation 10 and 11 as functions:
jp=2rnp
ra
(cos(α)−cos(α−θp
2np
)) + 2rnt
ra
(1 −cos θt
2nt
)
jt=2rnt
ra
(cos(α)−cos(α−θt
2nt
)) + 2rnp
ra
(1 −cos θp
2np
)
(12)
In the case that either npor ntare 0, the second term
which is the coupling effect can be eliminated. This allows a
direct algebraic solution for jpto θpand jtto θt. However,
generally, we can define a Jacobian that relates the partial
derivatives of joints qto the motor joints jpos :
˙
jpos =Ja(q) ˙q(13)
˙
jp
˙
jt="djp
dθp
djp
dθt
djt
dθp
djt
dθt#˙
θp
˙
θt(14)
Jashall be referred to as the Actuation Jacobian. The
Actuation Jacobian is a 2×2square matrix that is invertible
and can be found by taking the partial derivatives:
Ja(θp, θt) = "−r
rasin(α−θp
2np)r
rasin θt
2nt
r
rasin θp
2np
−r
rasin(α−θt
2nt)#(15)
Using the Actuation Jacobian to solve qgiven jpos, we
start with an initial estimate ˆqand determine an update to
minimise the error:
∆j=jpos −ˆ
j(ˆq)(16)
This error can be minimised by taking the inverse of the
Actuation Jacobian to get an update for the next estimate:
dˆq=J−1
a(ˆq)∆j(17)
After several iterations, ∆jconverges to 0or at least
towards a small value using real noisy data where estimate
ˆqbecomes the solution to qgiven jpos.
When working with mmodules, the tendons in module k
are affected by the previous configuration in module k−1.
This coupling effect can be identified by isolating the motor
effect of module kon module k−1. In the two-module
example, the proximal module angles θ1p, θ1tcan be solved
using the iterative updates in equation 17. If the distal joint
angles θ2p, θ2tare set as 0, the change in distal tendon
lengths in the proximal module can be computed and the
motor effect is solved using equation 11. This isolated motor
effect can be subtracted from the distal motor angles in
jpos allowing the uncoupled distal motor angles to be used
to solve the actual θ2p, θ2tusing the iterative method of
equation 17. The overall solution to qis achieved and is
fed into the forward kinematics to repeat the cycle of the
control loop.
IV. EXP E RI MEN TS
To validate the control of the instrument, SnakeRaven was
attached to the RAVEN II and teleoperation experiments
were conducted to measure the control error in the physical
system. To determine the true tool point, the position was
measured by the Northern Digital Inc. Aurora Electromag-
netic (EM) tracking system.
Fig. 5. Experimental setup of SnakeRaven and sensor locations on robot
A. Setup
SnakeRaven was assembled and inserted into the RAVEN
II adaptor before teleoperation. The integrated system was
calibrated manually by ensuring that it started from the con-
figuration where the instrument is perpendicular to the table.
Four EM sensors were attached with three at geometrically
known locations on a 3D printed mount near the RCM and
one at the tip of the end-effector as seen in the magnified
inlets of Fig. 5.
The end-effector sensor was placed where the fibre-optic
tool would have been but in the opposite direction for ease
in removal after use. The effect of magnetic interference was
observed near the motors for the RAVEN II so the sensor
mount positioned the sensors closer to the RCM as opposed
to the RAVEN II base. The tracking error for the position and
orientation of the tool was recorded during a teleoperation
task to bend towards the four corners in a rectangular path.
B. Results
From user observation, the robot was able to success-
fully respond and move the tip of SnakeRaven to reach a
commanded pose using the keyboard. Figure 6, in the left
graph, shows the comparison between the EM tracking the
end-effector (in blue), the forward kinematic pose (in red)
and the desired pose trajectories (in green). The absolute
error in the controller is plotted for the position in the top
right sub figure and the orientation in the bottom right sub
figure. The blue signal represents the error between the EM
tracking pose to the estimated forward kinematics pose while
the orange signal represents the control error in the inverse
kinematics controller. The control loop iterations per second
were measured to be approximately 536Hz while the RAVEN
II ROS publisher and subscriber updates came in at 1000Hz
and the EM tracking was about 20Hz. The absolute mean
error of the inverse kinematics controller in comparison to
the EM tracking of the tool is recorded in table II
V. DISCUSSION
Overall, the experiments showed that SnakeRaven could
successfully integrate with the RAVEN II and be teleoperated
TABLE II
END -EFF EC TOR CO N TRO L ABSOLUTE ERRO R
Controller EM Tracking
Position (mm) Orientation (°) Position (mm) Orientation (°)
0.3175 0.3577 3.774 8.8692
-50
-295 55
-45
-300
Y (mm)
X (mm)
60
-40
-305
SnakeRaven Rectangular Trajectory Results
65
Z (mm)
-35
-310
-30
-25
EM Tracker Trajectory
Forward Kinematics
Desired Trajectory
0 1000 2000 3000 4000 5000 6000
Iterations
0
2
4
6
8
10
12
14
Position Error (mm)
Position Error
EM error
Control error
0 1000 2000 3000 4000 5000 6000
Iterations
0
10
20
30
40
50
Orientation Error (degrees)
Orientation Error
EM error
Control error
Fig. 6. SnakeRaven teleoperation trajectory and control error
to follow a trajectory provided by an inverse kinematics
controller. Further comparison to a design in the literature
[18]–[21] in table III, shows that SnakeRaven stands out as
a smaller, more dexterous and customisable robot with closed
loop kinematic control.
Figure 6 showed that the inverse kinematics algorithm was
able to minimise the control error in response to keyboard
teleoperation commands. It also showed that the trajectory
from the EM tracking system is close to the estimated motion
of the robot but with some deviated bias.
Analysing the data collected, the cause for the observed
deviation can be a result of registration misalignment in
between the transformation from the EM device and the robot
base, calibration error, fabrication offsets or offsets in how
the sensors were placed. Not to mention the possibility of
some electromagnetic interference that could cause outliers
in the EM tacking trajectory. Keyboard teleoperation is also
an unideal form of teleoperation as it involves discontinuous
updates from an input buffer to the desired trajectory which
can cause some undesired performance such as the spike
at about 4500 iterations in fig. 6. Cable transmission in
both the RAVEN II and the snake-like manipulator can also
cause issues due to backlash which can also cause a bias
and or a fluctuation in position and orientation error in the
EM tracking. Nevertheless, these issues did not prevail the
controller as in practise the offsets are small enough for the
user to compensate manually during teleoperation.
For this setup, the controller suffices the objectives of
the teleoperation; however, for future work more sensing
capabilities would be required to compensate for these offsets
in more precise control and a better user interface.
TABLE III
DESIGN COMPARISON TO LITERATURE
Property SnakeRaven Design [18]–[21]
Shaft Diameter [mm] 4 8
Maximum Bending Angle (°) ±92 ±75
VI. CONCLUSION
In conclusion, a patient-specific low-cost 3D printable
snake-like manipulator was teleoperated as a new instrument
for the RAVEN II platform. This snake-like tool was mod-
elled to be parametric for patient-specific interventions in
fibre-optic knee arthoscopy requiring an inverse kinematic
controller that can steer the position and orientation of the
tip. A controller was designed and implemented on the
physical SnakeRaven prototype. The teleoperation system
successfully allowed the user to steer the tip at different
positions and orientations in a test teleoperation task. Control
error was minimised but physical deviations were present in
practise which was compensated over by manual teleopera-
tion. Further work would explore improving the accuracy of
the robot by using vision sensing to compensate these offsets
for autonomous control. The current robot, at this stage, has
not been tested by professional surgeons and further work to
improve the teleoperation experience from their feedback is
required.
The results also indicate the feasibility of integrating
customised snake-like instruments that can potentially be
optimised for different tasks or patient anatomies. An end-to-
end framework that enables automatic design of a bespoke
snake-like instrument to be integrated to the RAVEN II and
optimised for dexterous manipulation is being investigated.
REFERENCES
[1] A. Jaiprakash, W. B. O’Callaghan, S. Whitehouse, A. Pandey, L. Wu,
J. Roberts, and R. Crawford, “Orthopaedic surgeon attitudes towards
current limitations and the potential for robotic and technological
innovation in arthroscopic surgery,” Journal of Orthopaedic Surgery,
vol. 25, 2017.
[2] T. Lendvay, J. Rosen, and B. Hannaford, “44 telerobotics : Its future
in clinical application,” 2012.
[3] B. Hannaford, J. Rosen, D. C. W. Friedman, H. King, P. Roan,
L. Cheng, D. Glozman, J. Ma, S. N. Kosari, and L. W. White, “Raven-
ii: An open platform for surgical robotics research,” IEEE Transactions
on Biomedical Engineering, vol. 60, pp. 954–959, 2013.
[4] G. Barbash and S. Glied, “New technology and health care costs–the
case of robot-assisted surgery.” The New England journal of medicine,
vol. 363 8, pp. 701–4, 2010.
[5] J. Burgner-Kahrs, D. C. Rucker, and H. Choset, “Continuum robots
for medical applications: A survey,” IEEE Transactions on Robotics,
vol. 31, pp. 1261–1280, 2015.
[6] V. Vitiello, S.-L. Lee, T. Cundy, and G. Yang, “Emerging robotic plat-
forms for minimally invasive surgery,” IEEE Reviews in Biomedical
Engineering, vol. 6, pp. 111–126, 2013.
[7] C. Culmone, G. Smit, and P. Breedveld, “Additive manufacturing of
medical instruments: A state-of-the-art review,” Additive manufactur-
ing, vol. 27, pp. 461–473, 2019.
[8] J. P. Desai, J. Sheng, S. Cheng, X. Wang, N. Deaton, and N. Rah-
man, “Toward patient-specific 3d-printed robotic systems for surgical
interventions,” IEEE Transactions on Medical Robotics and Bionics,
vol. 1, pp. 77–87, 2019.
[9] P. Dario, C. Paggetti, N. Troisfontaine, E. Papa, T. Ciucci, M. Car-
rozza, and M. Marcacci, “A miniature steerable end-effector for
application in an integrated system for computer-assisted arthroscopy,”
Proceedings of International Conference on Robotics and Automation,
vol. 2, pp. 1573–1579 vol.2, 1997.
[10] L. Wu, A. Jaiprakash, A. Pandey, D. Fontanarosa, Y. Jonmohamadi,
M. Antico, M. Strydom, A. Razjigaev, F. Sasazawa, J. Roberts, and
R. W. Crawford, “Robotic and image-guided knee arthroscopy,” 2020.
[11] M. Kutzer, S. M. Segreti, C. Y. Brown, M. Armand, R. Taylor, and
S. Mears, “Design of a new cable-driven manipulator with a large open
lumen: Preliminary applications in the minimally-invasive removal
of osteolysis,” 2011 IEEE International Conference on Robotics and
Automation, pp. 2913–2920, 2011.
[12] R. Murphy, M. Kutzer, S. M. Segreti, B. C. Lucas, and M. Armand,
“Design and kinematic characterization of a surgical manipulator with
a focus on treating osteolysis,” Robotica, vol. 32, pp. 835–850, 2014.
[13] P. Francis, K. Eastwood, V. Bodani, K. Price, K. Upadhyaya, D. J.
Podolsky, H. Azimian, T. Looi, and J. Drake, “Miniaturized instru-
ments for the da vinci research kit: Design and implementation of
custom continuum tools,” IEEE Robotics Automation Magazine,
vol. 24, pp. 24–33, 2017.
[14] G. Wu, D. J. Podolsky, T. Looi, L. A. Kahrs, J. M. Drake, and
C. Forrest, “A 3 mm wristed instrument for the da vinci robot:
Setup, characterization, and phantom tests for cleft palate repair,” IEEE
Transactions on Medical Robotics and Bionics, vol. 2, pp. 130–139,
2020.
[15] X. Chu, H. W. Yip, Y. Cai, T. Y. Chung, S. Moran, and K. W. S. Au, “A
compliant robotic instrument with coupled tendon driven articulated
wrist control for organ retraction,” IEEE Robotics and Automation
Letters, vol. 3, pp. 4225–4232, 2018.
[16] T. K. Morimoto, E. W. Hawkes, and A. M. Okamura, “Design of
a compact actuation and control system for flexible medical robots,”
IEEE Robotics and Automation Letters, vol. 2, no. 3, pp. 1579–1585,
2017.
[17] L. Wu, R. Crawford, and J. Roberts, “Dexterity analysis of three 6-dof
continuum robots combining concentric tube mechanisms and cable-
driven mechanisms,” IEEE Robotics and Automation Letters, vol. 2,
pp. 514–521, 2017.
[18] C. A. Velasquez, H. H. King, B. Hannaford, and W. J. Yoon,
“Development of a flexible imaging probe integrated to a surgical
telerobot system: Preliminary remote control test and probe design,”
pp. 894–898, 2012.
[19] W. J. Yoon, C. Vel´
asquez, L. W. White, B. Hannaford, Y. S. Kim, and
T. S. Lendvay, “Preliminary articulable probe designs with raven and
challenges: Image-guided robotic surgery multitool system,” Journal
of Medical Devices-transactions of The Asme, vol. 8, p. 014505, 2014.
[20] C. Vel´
asquez, N. V. Navkar, A. Alsaied, S. Balakrishnan, J. Abinahed,
A. Al-Ansari, and W. J. Yoon, “Preliminary design of an actuated
imaging probe for generation of additional visual cues in a robotic
surgery,” Surgical Endoscopy, vol. 30, pp. 2641–2648, 2015.
[21] C. A. Velasquez, Y. S. Kim, T. S. Lendvay, B. Hannaford, and W. J.
Yoon, “Raven eyes around the instrument from modular axis sharing,”
International Journal of Control, Automation and Systems, 2019.
[22] E. Hunziker, “Articular cartilage repair: basic science and clinical
progress. a review of the current status and prospects.” Osteoarthritis
and cartilage, vol. 10 6, pp. 432–63, 2002.
[23] M. Leunig, A. Leunig, K. Bise, and A. Goetz, “Laser induced removal
of meniscus cartilage and tendines,” Lasers in Medical Science, vol. 7,
pp. 369–374, 1992.
[24] J. Lee, J. Kim, K. Lee, S. Hyung, Y. Kim, W. Kwon, K. S. Roh,
and J. Choi, “Modeling and control of robotic surgical platform for
single-port access surgery,” 2014 IEEE/RSJ International Conference
on Intelligent Robots and Systems, pp. 3489–3495, 2014.
[25] Y. Kim, S. Cheng, S. Kim, and K. Iagnemma, “A stiffness-adjustable
hyperredundant manipulator using a variable neutral-line mechanism
for minimally invasive surgery,” IEEE Transactions on Robotics,
vol. 30, pp. 382–395, 2014.
[26] Uw-Biorobotics, “uw-biorobotics/raven2,” May 2018. [Online].
Available: https://github.com/uw-biorobotics/raven2
[27] H. H. I. King, S. N. Kosari, B. Hannaford, and J. Ma, “Kinematic
analysis of the raven-ii(tm) research surgical robot platform,” 2012.
[28] S. Chiaverini, B. Siciliano, and O. Egeland, “Review of the damped
least-squares inverse kinematics with experiments on an industrial
robot manipulator,” IEEE Trans. Control. Syst. Technol., vol. 2, pp.
123–134, 1994.
