Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

Modeling and Simulation of a Tracked Mobile Inspection Robot

in MATLAB and V-REP Software
Submitted: 16th December 2016; accepted: 18th February 2017

Michał Ciszewski, Łukasz Mitka, Tomasz Buratowski, Mariusz Giergiel

resolution. Nayak and Pradhan proposed a robot for

DOI: 10.14313/JAMRIS_2-2017/11
horizontal and vertical pipes. It is designed to oper-
Abstract: ate in pipes with constant radius [12]. Its construc-
This paper presents modeling and simulation of a tracked tion allows to use only one actuator to pass through
mobile robot for pipe inspection with usage of MATLAB pipeline elbows. Sharma et al. proposed an inspection
and V-REP software. Mechanical structure of the robot robot with additional functionality, which is obsta-
is described with focus on pedipulators, used to change cles removal and pipe cleaning [13]. They have also
pose of track drive modules to adapt to different pipe tested a possibility of controlling the robot wirelessly.
sizes and shapes. Modeling of the pedipulators is shown Another approach was presented by ULC company
with application of MATLAB environment. The models [14]. They designed a robot, called Micro Magnetic
are verified using V-REP and MATLAB co-simulations. Crawler, capable of inspecting any object made of
Finally, operation of a prototype is shown on a test rig. steel. This compact robot uses magnetic attraction,
The robot utilizes joint space trajectories, computed with thus can move in any direction, even upside down, on
usage of the mathematical models of the pedipulators. ferromagnetic surfaces. A manually adjustable robot
chassis for variable pipe geometry was designed by
Keywords: tracked robot, pipe inspection, kinematics, Inuktun company [9]. The Versatrax robot can be set
robot adaptation, V-REP, MATLAB, co-simulation for motion inside of pipes of various diameters or can
move on horizontal surfaces.
The research shows that an automatic adaptation
of the robot’s motion unit to the pipe shape and diam-
1. Introduction eter has not been investigated by research teams. In
In this paper, modeling and simulation of a pipe this paper, an automatic adaptation system is shown
inspection mobile robot are presented. The work is in the context of modeling, simulation and prototype
focused on a motion unit of the robot that can adapt validation.
to various shapes and dimensions of pipelines. To re-
alize motion of the robot, modeling and calculations 2. Mechanical Structure
are performed in MATLAB software and simulations The mechanical structure of the robot is based on
are run in MATLAB, connected to V-REP robotic simu- two pedipulators that control pose of the track drive
lator. To validate the simulations, prototype testing in
an analogous environment is shown. V-REP (Virtual
Robot Experimental Platform) is a general-purpose
robot simulation framework, widely used in robotics
[4]. It is rich with functions, features and comprehen-
sive APIs for interfacing with virtually any application
or device. The most important attribute that affects
functionality of the simulator is a distributed architec-
ture, this means each simulation element can be con-
trolled in different ways. These ways include regular
Application Programming Interface (API) with avail-
able programming languages: Lua, C/C++; remote API
with available languages: Lua, C/C++, Python, Java,
MATLAB/Octave and Urbi or Robot Operating System
(ROS) interface which can advertise or subscribe to
ROS topics and offer ROS services.
Latest research in the field of pipe inspection ro-
bots has been made by some teams. Hansen et. al. pro-
posed a wheeled mobile robot intended for inspec-
tion and pipe mapping [8]. Their robot is designed for
movement in horizontal pipes and utilizes a forward Fig. 1. Pedipulator mechanism [7]: 1 – track drive
facing, fisheye camera. This setup allows to perform module; 2 – front driven ring; 5 – rear driven ring;
visual odometry and build pipe models with high 3, 6 – sprockets; 4 – front arm; 7 – rear driven arm

5
 Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

modules. One pedipulator consists of two coaxial, in-

dependently driven rings (2, 5), with axes of rotation
in the center of the robot body, as depicted in Fig. 1.
These rings are driven by servomotors and torques
are transferred by internal meshing gear transmis-
sions (3, 6). Arms (4, 7) are mounted to the rings by
revolute joints. Other sides of the arms are attached
to the track drive module (1). The rear arm (7) is
equipped with an additional servomotor.
By assembling two pedipulators to the robot body,
a  reconfigurable drive mechanism was constructed
that allows positioning of the track drives to allow
robot adaptation to various environments [2]. The
robot is equipped with six servomotors, responsible
for setting pose of the track drive modules, thus the
robot has 8 drives in total. Construction of the robot Fig. 2. Kinematic model of pedipulator: θr1 – rear ring
guarantees that it can operate in pipelines with ac- (5) rotation angle; θf1 – front ring (2) rotation angle; θr3
tive water flow, sewage pipes and other environments – rear arm (7) rotation angle with respect to track drive
where watertightness and dust protection is required. module (1); θr2, θf2– unactuated joints rotation angles

3. Kinematic Model of the Pedipulator

Kinematic modeling of the robot was divided into nism can be treated as planar, since all the revolute
two main parts. The first objective was to create a joints axes are parallel. In order to apply an approach
model of the robot motion on flat and inclined sur- widely used in robotics for open kinematic chains, the
faces. This model was described in [3]. The model, mechanism was divided into two planar manipulators
however, does not provide information on the adap- with two and three degrees of freedom that represent
tation of the robot’s motion unit to the environment, front and rear parts of the pedipulator. Later, it was
since motion is described for only one pedipulator assumed that the manipulators have to maintain po-
pose, dedicated for even surfaces. To complement sition of their end-effectors in one point, that would
the modeling approach, a mathematical model of the fulfil constraints of the closed kinematic chain.
pedipulators that set pose of the track drive modules For both manipulators, transformation matrices
was created. were created, using standard Denavit-Hartenberg no-
In general, the pedipulator structure for adapta- tation [11]. The transformation from end-effector to
tion of one track module is a closed kinematic chain the base coordinate system for the front (2-DOF) ma-
that consists of five links and six revolute joints, nipulator is presented by equations (1) and (2), and
among which three are actuated. Their angles of for the rear (3DOF) manipulator by equations (3) and
rotations are denoted by and in Fig. 2. The mecha- (4).

(1)

where: θf1, θf2 – rotation angles of joints 1, 2; af1, af2 – lengths of links 1, 2.

(3)

(4)

where: θr1, θr2, θr2 – rotation angles of joints 1, 2, 3; saf1, af2 – lengths of links 1, 2, 3.

6 Articles
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

(5)

(6)

Poses of the pedipulators for particular pipe sizes tion. The second step was to generate 5th order poly-
were determined with usage of the CAD model. With nomial trajectories for the front (2-DOF) manipulator
usage of the data, joint angular positions were used as between initial and final poses. Next, inverse kinemat-
input for initial and final poses for trajectory genera- ics problem was solved with usage of a numerical ap-
proach, based on the Jacobian pseudo-inverse for the
rear (3-DOF) manipulator to match position of end-
a effector of the front manipulator. In this case, ana-
lytical formulation of the inverse kinematics problem
gives 8 solutions for the entire pedipulator structure.
Therefore, additional conditions were used for planar
manipulators, as in [11]. Joint rotation ranges were
also limited, to comply with the mechanical structure.
Finally, oscillations of the manipulator, caused by
transitions through singular positions were eliminat-
ed. The calculations were done in MATLAB software
with Robotics Toolbox [5]. Thus, smooth trajectories
for transformation of the robot pedipulators were ob-
tained to realize required motion in the workspace.
Original construction of the pedipulator as a closed
kinematics chain ensures that it is sufficient to use
b three motors to set desired pose of the track drive
module. Rotation angles of the positioning drives dur-
ing transformations are shown in Fig. 3.
Initial validation of the model was performed in
MATLAB. The pedipulators positions were drawn for
every step of the trajectory. In Fig. 4, the kinematic
model of both pedipulators is depicted during trans-
formation from neutral pose (all servomotors are in
the middle of their operating range) to the pose that
allows the robot to move in horizontal pipes with in-
ternal diameter of Ø210 mm. The MATLAB visualiza-
tion of the pedipulators model shown in Fig. 4 proves
that the trajectory calculation algorithm is properly
designed for reconfiguration of the robot chassis. The
analyzed closed kinematic chain is highly dependent
Fig. 3. Pedipulator transformation – joint trajectories: on the path taken to attain desired pose, thus it is not
a – neutral pose to horizontal pipe ϕ300  mm, always possible to use arbitrary start pose and expect
b – neutral pose to vertical pipe ϕ242  mm satisfactory kinematics results for every goal pose.

Articles 7
 Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

a The first step of the simulation procedure was to

create a properly defined robot model in V-REP that
satisfies all mechanical constraints and can be simu-
lated in a way resembling real operation. The robot
model designed in Autodesk Inventor software was
imported in the V-REP environment. Next, the model
structure was defined in a way that allows creation
of joints, motors and ensures interaction with en-
vironment. To improve calculation efficiency, the
model was composed of visible bodies and dynami-
cally enabled bodies that were simplified using con-
vex hull decomposition. Overview of the robot model
is presented in Fig. 6. To improve visual aspects of
the simulation, the tracks were modeled using sepa-
rate segments that move on a predefined path (Fig.
b 6a). In contrast, for dynamic model, each track was
represented by five cylinders with rounded external
surfaces that provide steady contact with flat and
curved pipe surfaces, whilst assuring simulation re-
sults analogous to a tracked drive. This setup proved
to be the most reliable for various simulation cases.
Natural representation of a track as a set of segments
is computationally costly and may lead to simulation
instabilities due to rapidly alternating contact points
of the track treads. The pedipulator mechanisms were
modeled using bodies connected by rotary joints that
were set as actuated with position controllers, or not
actuated, according to the mechanical design.
The robot model created in V-REP was linked with
MATLAB using a remote API link. The link allows run-
Fig. 4. Kinematic model of two assembled pedipula- ning simulations directly from MATLAB, with simul-
tors during transformation (white color – free joints, taneous data transfer between both programs. The
hatched – actuated joints): a – neutral pose, b – pose mathematical model created in MATLAB was used to
for motion in vertical pipe with diameter ϕ 242  mm control positions of the servomotors to realize pedip-
ulator transformation trajectories and also for veloc-
ity control of the track drive modules of the robot 3D
model created in V-REP.
3. Robot Motion Simulations in V-REP and
MATLAB software
In order to validate the mathematical model and
trajectory calculations of the robot’s pedipulators, it
was necessary to use a simulation environment. As
previously stated, V-REP is a versatile and customiz-
able simulation platform that can be easily connected
with remote applications such as MATLAB and may
be used for various applications [10].
Due to the fact that the robot control algorithms,
presented in this paper are developed in MATLAB,
a remote API was used. Remote APIs, independently
of programming language chosen, offer four opera-
tion modes, designed for different tasks (Fig. 5). First
mode is a blocking function call that causes an API
client to wait for response of the simulator. The sec-
ond type is a non-blocking function call that could
be used only for sending data to the simulator, as it
does not wait for a server response. The third type
is data streaming, where a client sends message once
and a server sends reply regularly to the client. The
last mode is synchronous operation that enables syn-
chronization of each simulation step with API client.
In this mode, server holds execution of each simula-
tion step, until a trigger from client is sent, thus it is
usually the slowest mode of communication. Fig. 5. V-REP API overview [4]

8 Articles
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

a a

b
b

Fig. 7. V-REP model of pipe inspection robot: a – pipe

run, b – robot driving through pipe elbows in DN315
and DN250 pipes

Fig. 6. V-REP model of pipe inspection robot: a – visible

bodies, b – dynamically enabled bodies (light color)
with joints (dark color)

Initially, reconfiguration of the pedipulators was

verified by V-REP simulations of the robot model
placed on a dedicated support that provided unob- a b c
structed motion of the track drive modules. Next the Fig. 8. V-REP model of pipe inspection robot during re-
robot motion was tested in various environments, configuration in the reducer of DN315 pipe to DN250
such as flat surfaces and pipes with circular cross- pipe: a – robot in pipe DN315, b – robot in the reducer,
section. In order to check operation of the robot in c – robot after reconfiguration to pipe DN250
complex pipe structures, a dedicated V-REP scene
was prepared (Fig. 7a). The test run contained the fol-
lowing pipe segments: straight ϕ300  mm, 90° bend
in ϕ300 mm pipe (Fig. 7b), followed by straight seg-
ment, reducer from ϕ300  mm to ϕ242 mm, straight
ϕ242 mm segment followed by 30° bend (Fig. 7c) and
straight section.
A co-simulation of MATLAB and V-REP was pre-
pared to check if the robot would be able to traverse
the pipe run. Reconfiguration of the pedipulators was
performed automatically based on the mathematical
model and motion of the robot in pipe was realized
by manual control, with usage of a joystick connected
to a MATLAB/Simulink model. In Fig. 7b, the robot V-
REP model is presented during motion in 90° bend in
Ø300 mm pipe. At this stage of control system design,
operation of passing bends was performed with us-
age of teleoperation with visual feedback.
In Fig. 8 we can observe the robot during simula-
tion in a pipe reducer between Ø300 mm and Ø242
mm pipes. This process is realized automatically, with
track velocity adapted to pedipulators reconfigura-
tion process. With this approach, servomotors over-
load is avoided, that could arise from rapid transfer
through the reducer by excessive velocity, manually
set with teleoperated control. The usage of MATLAB
and V-REP co-simulation run in synchronous mode al- Fig. 9. V-REP and MATLAB simulation results com-
lowed to control the robot 3D model in a similar way parison for transformation from neutral position
as a real prototype, because application of mathemat- to pipe Ø300 mm: a – positioning drives angles,
ical models was tested and verified visually. b – absolute positioning error V-REP vs MATLAB

Articles 9
 Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

a b

Fig. 10. V-REP and MATLAB simulation results com-

parison for transformation from pipe Ø300 mm
to pipe Ø242 mm: a – positioning drives angles,
b – absolute positioning error V-REP vs MATLAB
Fig. 11. Tests of the prototype: a – test rig, b – the
robot in DN250 pipe (Ø242 mm internal diameter),
Results of the simulations are presented in Fig. 9 c – the robot after completion of the test run
and Fig. 10. We can see that the mathematical model
trajectories from MATLAB and simulation results
from V-REP coincide well and the absolute position- manual operation of track drives by velocity setting
ing errors of pedipulators drives do not exceed 0.6° and automatic transformation of the robot’s pedipu-
for the analyzed cases. The errors are caused mainly lators. Motion in the test rig included the following
by joint position tolerances of the Bullet dynamics en- steps: drive into the pipe, traverse 90° bend in larger
gine, used for the simulation, especially when closed pipe, get to pipe narrowing, reconfigure motion unit
kinematic chain is analyzed. This errors are satisfac- to smaller pipe diameter, get into smaller pipe (Fig.
tory and do not affect proper transformation of the 11b), pass to next bend, traverse 30° bend in smaller
pedipulators between desired poses. pipe, get out of the test rig. The performed test com-
V-REP and MATLAB co-simulations proved to be plied with simulation results and the robot completed
very efficient testing tools for motion analysis of the the experiment successfully (Fig. 11c).
robot models that allowed optimization of pedipula- In straight pipe segments, motion of the robot
tors transformation trajectory calculation algorithm without slip was observed, however negotiation of
and development of control system for the robot pro- pipe bends implied excessive slip of track drives that
totype. were controlled manually. Application of an auto-
matic velocity control during negotiation of bends
4. Tests of the Prototype would provide smoother transitions. The most dif-
For the testing of the robot’s ability to adapt its ficult operation was to drive through pipe reducer
pedipulators poses and assessment of prototype mo- from DN315 mm to DN250 mm. In this case, simulta-
bility inside pipes, a test rig was prepared. It consisted neous reconfiguration of the robot’s pedipulators and
of straight pipes of nominal diameter DN315 mm and the tracks forward motion would be optimal, but the
DN250 mm, pipe reducer connecting two pipes and functionality was not available in the current version
two bends, one 90° bend of diameter DN315 mm and of control system of the prototype.
second 30° bend in pipe of diameter DN250. The pipe
segments were assembled as presented in Fig. 11a, 5. Conclusions and Future Work
similarly to the simulation scene presented in Fig. 7. In this paper, mathematical modeling, simulations
The objective of the test procedure was to verify and tests of a pipe inspection robot with an adaptable
if the robot would be able to traverse the rig. Initially, motion unit were shown. Co-simulation prepared in
motion unit of the prototype was reconfigured to the VREP simulation software with MATLAB/Simulink
pose for motion in DN315 mm pipe, with usage of the computation environment proved to be very efficient
trajectories for positioning servomotors, computed tools for testing control system design for the inspec-
in MATLAB. Next, the robot was placed at the entry tion mobile robot, featuring two pedipulators with
of the pipeline. Control strategy of the robot included closed kinematic chains. The pedipulators transfor-

10 Articles
Journal of Automation, Mobile Robotics & Intelligent Systems VOLUME 11, N° 2 2017

mation trajectories, generated with usage of math- [6] J. Giergiel, M. Giergiel, T. Buratowski, M. Cisze-
ematical models, were also verified using a prototype wski, „Pedipulator’s mechanism for positioning
on a dedicated test rig analogous to simulation en- of a track drive module, esp. for mobile robots”,
vironment. The preliminary prototype tests allowed Patent application (description of the invention)
validation of the pedipulators reconfiguration but did no. PL 406656, 2015 (in Polish).
not include verification of a complete control system [7] P. Hansen, H. Alismail, P. Rander, B. Browning,
for robot motion. “Visual mapping for natural gas pipe inspec-
Future work would focus on software implemen- tion”, Int. J. Rob. Res., vol. 34, 2015, 532–558.
tation for the inspection robot, including a vision DOI: 10.1177/0278364914550133.
system that would serve as an aid for autonomous [8] J. Khalilov, A.T. Kutay, “Interfacing Matlab/Simu-
reconfiguration inside pipelines, depending on the link with V – REP for A Controller Design for
recognized geometry. Moreover, the robot would be Quadrotor”, Int. J. Eng. Res. Rev., no. 3 , 2015,
simulated during motion and adaptation in vertical 42–49.
pipes. An upgraded electronic board and enhanced [9] K. Kozłowski, P. Dutkiewicz, W. Wróblewski,
software would allow control of the track drive clamp Modelowanie i sterowanie robotów (Control and
forces exerted by pedipulators extension during mo- modelling of robots), 1st ed., PWN: Warszawa,
tion of the robot inside of vertical pipelines. 2012. (in Polish)
[10] A. Nayak, S.K. Pradhan, “Design of a new in-
pipe inspection robot”, Procedia Eng., no. 97,
AUTHORS 2014, 2081–2091. DOI: 10.1016/j.pro-
Michał Ciszewski* – Department of Robotics and Me- eng.2014.12.451.
chatronics, AGH University of Science and Technology, [11] Y. Sharma, K. Deepak, P. Kumar, A. Chauhan,
Cracow, Poland. E-mail: [email protected]. “Blockage removal and RF controlled pipe in-
spection robot (BRICR)”, Int. J. Electron. Telecom-
Łukasz Mitka – Department of Robotics and Mecha- mun., vol. 4, 2015, 62–68.
tronics, AGH University of Science and Technology, [12] Coppelia Robotics, V-REP Virtual Robot Experi-
Cracow, Poland. E-mail: [email protected] mental Platform, 2016, http://www.coppeliaro-
botics.com
Tomasz Buratowski – Faculty of Mechanical Engi- [13] Inuktun, Inuktun crawler vehicles, 2015, http://
neering and Robotics AGH University of Science and www.inuktun.com/crawler-vehicles
Technology, Poland. E-mail: [email protected] [14] ULC, ULC’S Micro-magnetic crawler, 2014,
http://ulcrobotics.com
Mariusz Giergiel – Faculty of Mechanical Engineer-
ing and Robotics, AGH University of Science and Tech-
nology, Poland. E-mail: [email protected]

*Corresponding author

REFERENCES

[1] S.R.S. Buss, Introduction to inverse kinematics with

jacobian transpose, pseudoinverse and damped
least squares methods, Univ. California, San Diego,
Typeset Manuscr., vol. 132, 2009, 1–19.
DOI: 10.1016/j.neuroscience.2005.01.020.
[2] M. Ciszewski, T. Buratowski, M. Giergiel, P. Mal-
ka, K. Kurc, “Virtual Prototyping, Design and
Analysis of an in-Pipe Inspection Mobile Robot”,
J. Theor. Appl. Mech., vol. 52, 2014, 417–429.
[3] M. Ciszewski, M. Waclawski, T. Buratowski, M.
Giergiel, K. Kurc, “Design, Modelling and Labo-
ratory Testing of a Pipe Inspection Robot”, Arch.
Mech. Eng., vol. 62, 2015, 395–408.
DOI: 10.1515/meceng-2015-0023.
[4] P. Corke, Robotics, Vision and Control: Fun-
damental Algorithms in MATLAB, Spring-
er, 2011, ISBN 978-3-642-20144-8
DOI: 10.1007/978-3-642-20144-8
[5] I. Dulęba, M. Opałka, “A comparison of Ja-
cobian-based methods of inverse kinemat-
ics for serial robot manipulators”, Int. J. Appl.
Math. Comput. Sci., vol. 23, 2013, 373–382.
DOI: 10.2478/amcs-2013-0028.

Articles 11