You are on page 1of 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/253585277

MOBILE ROBOT LOCALIZATION IN A MANUFACTURING ENVIRONMENT

Conference Paper · October 2008

CITATIONS READS

5 1,714

3 authors:

Zoran Miljković Najdan Vuković


University of Belgrade-Faculty of Mechanical Engineering University of Belgrade
104 PUBLICATIONS   1,573 CITATIONS    38 PUBLICATIONS   741 CITATIONS   

SEE PROFILE SEE PROFILE

Bojan Babic
Institute of Biotechnology
52 PUBLICATIONS   864 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chaotic metaheuristic algorithms for learning and reproduction of robot motion trajectories View project

Deep Machine Learning and Swarm Intelligence-based Optimization Algorithms for Control and Scheduling of Cyber-Physical Systems in Industry 4.0_MISSION4.0 View
project

All content following this page was uploaded by Najdan Vuković on 22 April 2014.

The user has requested enhancement of the downloaded file.


MOBILE ROBOT LOCALIZATION IN A MANUFACTURING ENVIRONMENT

Z. Miljković1, N. Vuković1, B. Babić1


1. The University of Belgrade, Faculty of Mechanical Engineering, Production Engineering Department,
Kraljice Marije 16, 11120 Belgrade 35, Republic of Serbia

ABSTRACT
This paper deals with the integration of Intelligent Mobile Robots (IMR) into the
advanced manufacturing paradigm known as Intelligent Manufacturing Systems
(IMS) and analyses some of the most important issues facing this advanced
integration. The paper introduces a localization problem of mobile robots and
provides a simulation procedure carried out in Matlab® environment. The presented
software application (the source code) provides a chance for modelling a localization
problem on the real shop floor. Finally, the first step towards an experimental setup
using the LEGO® MindstormsTM NXT robotic kit is presented. Although this approach
based on simulation may be seen as neglecting essential issues by introducing
assumptions common in real problem modelling, it still provides a chance for proper
development of the model.

KEYWORDS: Intelligent Manufacturing Systems, Material Transport System,


Intelligent Mobile Robot, Localization, Simulation

1. INTRODUCTION
At the beginning of the 21st century manufacturing is more closely than ever related to fast
growing market requirements and intensively coupled with diverse customer demands. The
increasing complexity of products and growing tendency for delivery time cutting as well as the
need for “make to order“ rather than “make to stock“ manufacturing, imposes newly developed
solutions able to tackle with these sophisticated issues. New methods, fast growing research
fields, design principles and newly developed and defined paradigms, guarantee improvement
of the existing technology as well as the quality of every day life.
Intensive research in the field of robotics has resulted in a great number of robots able to
perform complex and sophisticated tasks they had been previously designed to do. Throughout
years robotics has achieved a number of important great successes in various fields of
application such as manufacturing, museum touring, cargo handling etc. However, one of the
greatest successes to date is in the world of industrial manufacturing where industrial
manipulators are able to move with great speed and accuracy performing all sorts of tasks, such
as welding, painting, cutting etc. /8/. Needless to say, implementation of industrial robots for
manufacturing purposes is a standard for highly-developed companies.
On the other hand, implementation of Intelligent Mobile Robots (IMR) in manufacturing systems
/10/ is still a challenge for the research community. Operating on the shop floor, as a component
of material transport system, an IMR would need a particular kind of behaviour exclusively
developed for these purposes. Having these facts in mind, this paper analyses integration of
IMR into advanced manufacturing paradigms such as Intelligent Manufacturing Systems (IMS).

2. MOTIVATION AND PRELIMINARIES


More than one million industrial robots have been installed worldwide executing various manu-
facturing tasks /10, 11/. However, the major disadvantage of industrial manipulators is lack of

Proceedings of the 3rd International Conference on Manufacturing Engineering (ICMEN), 1-3 October 2008, Chalkidiki, Greece
Edited by Prof. K.-D. Bouzakis, Director of the Laboratory for Machine Tools and Manufacturing Engineering (ΕΕΔΜ),
Aristoteles University of Thessaloniki and of the Fraunhofer Project Center Coatings in Manufacturing (PCCM),
a joint initiative by Fraunhofer-Gesellschaft and Centre for Research and Technology Hellas, Published by: ΕΕΔΜ and PCCM

485
mobility. They have restricted workspace /6/ due to the fact that they are bolted to a fixed posi-
tion. In other words, if there is a need for a robot to change its work volume /6/ or even the
shape of the initial work volume for the purposes of a particular task, then the robot will not be
able to perform the task successfully. In recent years one may observe ongoing changes we
have been witnessing in the world of manufacturing, i.e. the dominant task of industrial robots
has shifted from Pick and Place to Fetch and Carry tasks /10/, increasing the need for mobile
industrial robots. The experimental setup which may be understood as an attempt to solve this
transition of the main task of industrial robots may be seen in /11/ where the four Degrees of
Freedom robotic arm was implemented on a wheeled mobile platform. It should be pointed out
that this fusion represents one of the possible applications of mobile robots in a manufacturing
environment.
Another aspect which should not be overlooked is correlated to problems of dominant industry
in some European countries, where the majority of enterprises can be classified into Small and
Medium Sized Enterprises (SME). As a general rule of thumb, these companies employ less
then 500 employees. One of the major flaws these companies suffer of is restricted “man-
power” capabilities resulting in partially implemented advanced technologies. For instance, fully
integrated concept of Computer Integrated Manufacturing (CIM) includes constant investments
in hardware and software. Therefore, advanced manufacturing paradigms could not be applied
thoroughly. One may say that CIM concept was exclusively developed for wealthy corporations
willing and able to invest into the better future. However, newly developed paradigms such as
Intelligent Computer Integrated Manufacturing (ICIM) and particularly Intelligent Manufacturing
Systems (IMS) could help SMEs to survive market requirements. Having this in mind, one may
notice the concept of flexibility as one of the most important concepts in the manufacturing
technology of the 21st century. The ability to adapt to market requirements, to manufacture and
finally to deliver in time is the essential capability. Therefore, this paper deals with the
localization of a mobile robot in a manufacturing environment in order to speed up the transport
system and consequently the manufacturing process itself.

2.1 Levels of integration in Computer Integrated Manufacturing Systems


Progressive manufacturing paradigms like Computer Integrated Manufacturing and Intelligent
Manufacturing Systems (IMS) may ensure the design of complex, sophisticated and more inte-
grated manufacturing environments developed on the basis of modern software architectures
and information technologies /2,8,10,11/. Integration of mobile robots in manufacturing should
be carried out after thorough analysis of all possible advantages and disadvantages which may
occur and on these bases one may consider a fulfillment of the initial assumption – Intelligent
Mobile Robots can be implemented into a Manufacturing Environment. The analysis should
point out weather IMR is to be integrated into the existing manufacturing environment or a new
environment should be designed from scratch. Consequenlty, this inference will provide
important information necessary to establish an appropriate integration process. It is a well
known fact that there are three main integration levels in CIM /2,10/:
• Level of mechanical integration – this level of integration is a hardware structure of Flexible
Manufacturing Systems (FMS), in other words FMS(M ∧ C ∧ IR), where M represents the
machine tool, C the computer and IR industrial robot.
• Level of comunication integration - integration of various control units using LAN (Local Area
Network) with machine tools on the shop floor, or symbolically: CIM(CAD ∧ CAM ∧
CAPP ∧ CAQ ∧ PPS ∧ FMS(M ∧ R ∧ IR)), where CIM stands for Computer Integrated Manu-
facturing, CAD for Computer Aided Design, CAPP for Computer Aided Process Planning,
CAQ Computer Aided Quality Control, and finally, PPS includes the use of computer aided
systems for organizational planning, controlling and monitoring of all processes.
• Integration using knowledge – this is the highest level of integration and assumes
implementaton of Artificial Intelligence (AI) and Knowledge engineering (KE), or symbolically:
IMS(AI ∧ CIM(CAD ∧ CAM ∧ CAPP ∧ CAQ ∧ PPS ∧ FMS(MA ∧ R ∧ IR), where acronyms
stand for already introduced concepts.

486 3rd ICMEN – 2008


Based on previous levels of integration one may easily deduce that if IMR is to be implemented
in a manufacturing environment then IMR should be properly integrated by applying three main
levels of integration in CIM systems. The first integration level defines the function of an
intelligent mobile robot in a manufacturing environment. This aspect of integration is described
in the subsequent part of the paper. The second level of integration presribes communication
among machine tools, including the mobile robot. Finally the highest level of integration is the
third level, defining how to integrate the knowledge. This integration level, to some extent,
introduces the Intelligent Manufacturing System as a paradigm in manufacturing and associated
technology as well.

2.2 Material Transport System


Material Transport System (MTS) is a significant component in manufacturing processes. How-
ever, it should be stressed that MTS is a part of a much broader concept known as Material
Handling (MH). MH is defined by the Material Handling Industry of America as ’’the movement,
storage, protection and control of materials throughout the manufacturing and distribution proc-
ess including their consumption and disposal’’ /8/. As one may easily infer, material handling
process must be performed safely, efficiently, at low cost and without damage to the goods. Al-
though often being overlooked in companies, material handling is an important issue in produc-
tion since the cost of material handling is a significant portion of the total production cost aver-
aging around 20% of the total manufacturing labor cost /8/. Having this in mind, it is easily un-
derstood why this hard problem should be in the focus of a research community, side by side
with other quintessential problems in a manufacturing environment. Therefore, this paper aims
to show the necessity of mobile robot integration as a component of MTS and consequently of
IMS.

2.3 Intelligent Mobile Robots


A manufacturing environment is extensively complex and inherently unpredictable. With every
new product being designed and manufactured the complexity of manufacturing technology in-
creases. On the other hand, associated manufacturing technology needs to be appropriately
developed in order to meet the prescribed requirements. Therefore, Material Transport System
as a component of associated manufacturing technology should be adequately developed as
well. It is easily understood that dynamical characteristics of a manufacturing environment im-
pose particular abilities a mobile robot should have if it is to operate on the shop floor efficiently,
accurately and successfully. Consequently, an IMR needs to adapt itself to everlasting changes.
Having this in mind, it is absolutely clear that an IMR should be intelligent, at least to some ex-
tent /11/. It is a well known fact that intelligent agents consist of the following basic building
blocks /7/: the sensing ability, the perceptive ability, the knowledge acquisition ability, the learn-
ing ability, the inference ability, the decision making ability and the acting ability.

3. PROBLEM STATEMENT
Unlike conventional industrial robots with a constant shape of the workspace, a mobile robot
may travel around the shop floor flexibly applying its talent wherever it is needed /14/. This ad-
vantage of the mobile robot technology should be exploited as much as possible. Mobile plat-
forms, representing a fusion between industrial robots and mobile robots, mentioned earlier in
the paper, may improve abilities of industrial robots and mobile robots as well. However, com-
paring these two, at first glance incomparable machines, may result in wrong conclusions.
Therefore, it needs to be stressed that both of these advanced machines have their own advan-
tages and disadvantages, so the fundamental idea of this paper is not to glorify mobile robots
but to point out that merging possibilities are vast and should be exploited as much as possible.
As it has already been said, intelligence is one of the building blocks of the mobile robot tech-
nology and, if achieved, it increases the robot’s selfawareness. Having the ability to adapt to
environment changes is a rather important capability of an intelligent system. However, the last

Robots in Manufacturing Environments 487


fact imposes the following: if a mobile robot is to operate in such a dynamic environment, like a
manufacturing environment, it must be able to determine its position and orientation in the envi-
ronment. Therefore, at each and every moment a mobile robot should provide answers to the
following questions /12,14/: Where am I?, Where have I been?, Where am I going?, and finally,
What’s the best way there? The first two questions fall in the realm of mobile robot localization.
Mobile robot localization is the problem of determining the pose (position and orientation) of a
robot relative to a given map of the environment /15/, and quite often is referred to as the pose
estimation problem. The third and the fourth questions are related to planning and control ability
of a mobile robot but those issues will not be analysed in this paper.
State estimation algorithms based on Kalman filter /9/ are a common approach for linear sys-
tems. The Kalman filter is a base of recursive state estimators called Gaussian filters. It is a well
known fact that Gaussian filters approximate function by its statistical moments, mean and co-
variance /3, 15/. In the same manner the Kalman filter represents beliefs by the first and the
second moment. However, accuracy of these algorithms when applied to nonlinear problems
decreases rapidly. Therefore, throughout years a number of possible solutions have been ad-
vised /3, 5, 15/ in order to improve the basic formulation and implementation. In this paper an
Extended Kalman Filter (EKF) /5, 15/ algorithm will be used for the localization purposes of a
mobile robot in a manufacturing environment. The formulation of EKF algorithm for the state
transition model is given by:
xt = g(ut,xt-1) + εt (1)
whereas the measurement model is given by:
zt = h(xt) + δt (2)
ut are controls, xt is robot’s pose at the time t, and xt-1 is robot’s pose at the time t-1. Function
g(ut,xt-1) represents state transition model, and h(xt) is the measurement probability. Parametars
εt and δt are random variables and represent noises of models. Common practice is to assume
these noises as Gaussian distribution with zero mean and covariance matrices Rt and Qt for
random variables εt and δt, respectively. The basic idea of the EKF approximation is in lineariza-
tion of a state transition model and measurement model via Taylor expansion. The Extended
Kalman Filter algorithm can be seen in Table 1. Algorithm “starts“ with the input of the mean μt-1,
and the covariance matrix ∑t-1, at the time instant t-1, control ut, and measurement zt at the time
instant t. The prediction step of EKF algorithm is defined in lines 2 and 3 where the a priori
mean and covariance matrix are calculated. Line 4 defines the Kalman gain. If one pictures the
Kalman gain as a scalar quantity then one may notice the gain is proportional to the state
prediction variance and inversly proportional to the innovation variance /3/. It specifes the
degree by which the measurement is incorporated into the new estimate /15/. Measurement
inovation is given in line 5 by the following term zt − h( µt ) . The correction step can be seen in
line 5 and line 6, where updated, a posteriori, mean and covariance are defined.

Table 1: The Extended Kalman Filter algorithm /15/.


1. ALGORITHM Extended _ Kalman _ filter ( µ t −1 , Σ t −1 , u t , z t );
2. µ t = g (u t , µ t −1 )
3. Σ t = Gt Σ t −1GtT + Rt
4. K t = Σ t H tT ( H t Σ t H tT + Qt ) −1
5. µ t = µ t + K t ( z t − h( µ t ))
6. Σ t = ( I − K t H t )Σt
7. return µ t , Σ t

488 3rd ICMEN – 2008


3.1 Motion model
In general, there are two models developed for the determination of posteriors of robot’s pose
over poses. The first model is based on odometry measurements. The word odometry is formed
of two Greek words, hodos - meaning journey, and metros meaning length. Odometry is known
as dead reckoning as well (derived from deduced reckoning). In other words, this concept
provides information regarding how far a vehicle (in general) has travelled based on rotation of
its motors, shafts encoders etc. The second model is based on translational and rotational
velocities of the robot, which may be defined as controls. Thourough analysis of mobile robot
kinematics can be seen in /14/ while the probabilistic models of mobile robot motion are
presented in /15/. For simulation purposes a velocity based model of differential drive mobile
robot was introduced and may be seen in Figure 1.

Figure 1: The model of differential drive mobile robot used in simulation.

The mathematical formulation of a state transition model (motion model) in discrete form is:

⎧ x(k + 1) ⎫ ⎡ x(k ) + v(k )Δt cos(θ (k ))⎤


⎪ ⎪ ⎢ y (k ) + v(k )Δt sin(θ (k )) ⎥
⎨ y (k + 1)⎬ = ⎢ ⎥ (3)
⎪θ (k + 1)⎪ ⎢⎣ θ (k ) + ω (k )Δt ⎥⎦
⎩ ⎭
and the associate control vector is given by u(k) = {v(k) ω(k)}T, where v(k) is the velocity and
ω(k) is the angular velocity at k-th time step. As usual, Δt is the time interval and robot pose is
given by p = {x y θ}T knowing that a mobile robot is operating in the plane.

3.2 Sensor model


The sensor model of a mobile robot perception developed for simulation purposes is based on
feature based maps /15/. Basically, a feature is a distinct object in the physical world /15/ fully
recognizable for a robot. In general, doors, corners, walls etc. are defined as features and it is
common to call them landmarks. The basic idea goes like this: each landmark has a unique and
distinguishable characteristic called signature. Based on these facts, the feature vector is de-
fined as /15/:
f ( zt ) = { f t1 , f t 2 ,... f t n } = {[ rt1 φt1 st1 ]T , [ rt 2 φt2 st2 ]T ,..., [rt n φtn stn ]T } (4)

where rt n stands for range, measured from the robot position to the landmark position, φtn is
bearing relative to the robot position, and finally stn represents the signature of the observed

Robots in Manufacturing Environments 489


landmark. zt is measurement, f(zt) stands for the feature vector, n is the number of the perceived
feature and index t is the time instant the measurements are taken. The key idea behind this
innovative approach is reduction of computational complexity. The sensor model is easily de-
rived using the basic geometric laws but with the addition of Gaussian noise for range, bearing
and signature respectively e.g.

⎛ rt n ⎞ ⎛⎜ (m jx − x) 2 + (m jy − y ) 2 ⎞⎟ ⎛⎜ ε σ 2 ⎞

⎜ n⎟ r
⎜ φt ⎟ = ⎜ a tan 2(m jy − y, m jx − x) − θ ⎟ + ⎜⎜ ε σ 2 ⎟

(5)
⎜ s n ⎟ ⎜⎜ sj

⎟ ⎜ε 2
φ

⎝ t ⎠ ⎝ ⎠ ⎝ σs ⎠

where mjx and mjy represent Cartesian coordinates of landmarks, ε σ 2 , ε σ 2 and ε σ 2 are noises
r φ s

modelled with Gaussians. For thorough analysis refer to /15/.

4. SIMULATION RESULTS

In order to test a localization procedure of a mobile robot in a manufacturing environment, a


simulation has been performed in Matlab® and the results of one simulation will be presented.
For these purposes a source code was written, tested and implemented. The source code can
be seen in Table 2.

Table 2: The core of Matlab® source code.


%======================================================================
while kk <= kkmax
d = compute_distance(pose, map(kk,:));
[phi,w,lm] = compute_direction(pose,map(kk,:),dt);
while d > eps
dtc = dtc + dt;
dto = dto + dt;
pose = motion_model(pose, [v w], dt, add_noise);
[pose_pred,C_]= EKF_predict(pose_pred,C_,v, w, M, dt);
if dto >= dt_observations
VL = observable_landmarks(pose, map, maxR);
VLNo = size(VL,1);
if isempty(VL)
pose_update = pose_pred;
else
[z] = observations(pose, VL, add_noise);
[zp,Hp,M_, Q_] = measurements(pose_pred, VL,M,Q)
iv = innovation_vector(z,zp);
[pose_update,C_,K,up,prev] = EKF_update(pose_pred,...
C_, iv, Hp, Q_ );
end
else
continue
end
end
end
%======================================================================

In Table 3 values of some parameters are given. These parameters are necessary for the
initialization of mobile robot localization simulation in a manufacturing environment.

490 3rd ICMEN – 2008


Table 3: Values of parameters.
Translational Angular Range Bearing
Time Translational Angular Perceptual measurement
velocity velocity measureme
interval velocity velocity range noise
noise noise nt noise
dt v ω Rmax σφ
σv σω σr
[s] [m/s] [rad/s] [m]
[m/s] [rad/s] [m] [rad]
0.01 0.5 0 0.5 0.1 0.349 0.03 0.1745

The algorithm starts by introducing the map of a manufacturing environment, i.e. map variable in
Table 2 which defines the position of machine tools on the shop floor in Cartesian coordinates.
In the main while loop the mobile robot travels to each and every machine tool constantly
updating the distance to the next machine (user-defined function compute_direction). There are
seven distinguishable features on the map representing the machine tools. Each and every
feature is recognizable helping the robot to localize itself. A mobile robot is to travel between
machines transporting materials and other goods needed for manufacturing processes on
machine tools. Figure 2 presents the four positions of the mobile robot while fulfilling the
prescribed task. In Figure 2 (a) one may see the initial position of the robot represented with the
following pose vector xt0= {0 0 0}T. The dotted line in Figure 2 (a) to (d) represents the path the
robot should travel. Seven machine tools are represented with circles. The circle with the
greatest diameter in Figure 2 is robot’s perceptual space, in other words the maximum range of
its proximity sensors e.g. laser range finders. When a mobile robot perceives a particular feature
(or features in gen eral, refer to Figure 2 (d)) it is assumed that the robot instantaneously and
undoubtedly knows the identity of the feature. Having this in mind, the localization algorithm
b a s e d o n E x t e n d e d

(a) (b)

(c) (d)
Figure 2: Simulation results of EKF localization algorithm.

Robots in Manufacturing Environments 491


Kalman Filter is able to determine the robot’s pose by incorporating the most recent measure-
ment into the belief. By knowing the features identity (defined by feature’s signature) the robot
updates its belief, hence improving the knowledge of its whereabouts. An interesting phenom-
ena occurs in Figure 3 where the most important characteristic of localization algorithm is pre-
sented. Namely, the mobile robot is travelling the longest segment of the defined path. How-
ever, while on the path due to inherent errors in the controls and sensors, modelled by uncer-
tainty in the motion model and sensor model, the mobile robot has lost track of its whereabouts.
Hence, Figure 3 presents two mobile robots, the “real” one and robot’s belief of its pose. In this
particular case, one may see the belief as a result of EKF based localization algorithm. Conse-
quences of this phenomenon could be devastating, since the robot’s belief is not consistent with
the position of the real robot. This problem may result in a situation well known as kidnapped
robot problem /15/, which represents a significant deviation of localization algorithm’s result from
the real robot’s pose. This particular issue is a common problem in real world implementation of
mobile robots. A typical kidnapped robot problem can be seen in Figure 3. However, having
perceived machine tools number four and number five (Figure 2 (c)), the robot successfully up-
dates its belief and manages to determine its pose.

Figure 3: Mobile robot while travelling the longest segment on the defined path.

5. DISCUSSION

One may notice extreme values for parametars given in Table 3. Namely, motion noises σv and
σω (noises for translational and angular velocity of the mobile robot respectively) and
measurement noises σr and σφ (range and bearing noises) are much higher than expected,
even for an actual manufacturing environment. This assumption results in the following manner:
the mobile robot pose estimation error is approximately ±5 [cm] for Cartesian coordinates and
20÷30º for orientation while travelling the course of ≈10 [cm]. Knowing the size of the simulated
robot (the diameter of the robot is 10 [cm]) one may argue about the usefulness of the localiza-
tion algorithm. Having a 20% error in the control system is absolutely impossible but the main
idea behind the simulation is in testing the robustness and accuracy of the presented localiza-
tion algorithm.

5.1 Towards Experimental Setup


In this paper, a brief introduction to an ongoing research at the University of Belgrade - Faculty

492 3rd ICMEN – 2008


of Mechanical Engineering is presented. The paper provides the definition of a localization prob-
lem, simulation procedure, Matlab® source code and obtained simulation results. However, be-
sides simulation of a mobile robot localization in a manufacturing environment one of the goals
of an ongoing research is to develop an experimental environment that resembles a particular
manufacturing environment. Proper implementation of the source code fulfilling its task while
localizing the “real” mobile robot operating in the experimental setup, would test the source code
to its limits.
Therefore, in Figure 4 an experimental mobile robot built on LEGO Mindstorms™ NXT technol-
ogy is presented. Differential drive kinematic configuration /14/ was chosen. The robot is to be
controlled via notebook PC and by using Matlab®. Needless to say that the localization source
code, which has been presented earlier in the paper, will be a part of a software being written
for these purposes.

Figure 4: Experimental setup.

6. CONCLUSION

This paper presents an ongoing research at the University of Belgrade - Faculty of Mechanical
Engineering. The research deals with integration of Intelligent Mobile Robots in a Manufacturing
Environment as a component of an advanced manufacturing paradigm known as Intelligent
Manufacturing Systems. For these purposes, an analysis of this progressive integration is
presented. Based on these results it was concluded that a mobile robot should be integrated
through mechanical, communication and knowledge levels of integration into a manufacturing
environment. Proper integration into an existing or a newly developed environment would
enable a mobile robot to fulfill tasks it had been designed to do. Finally, it has been pointed out
that a mobile robot should be intelligent in order to operate in an unpredictable environment
such as the shop floor of a manufacturing environment.
Having this in mind, a characteristic issue in the field of mobile robotics was introduced, namely,
the localization problem. If a mobile robot is able to localize itself relative to the given map of the
environment, then one may deduce that a particular mobile robot is an intelligent creature. Hav-
ing the ability to estimate its pose in such a dynamic environment is one of the most important
issues facing full integration. On these bases, a localization algorithm based on Extended Kal-
man Filter was introduced and simulation results were presented. The model of a differential
drive mobile robot with three degrees of freedom was introduced and, based on this model,
simulation was performed in Matlab® environment. For these purposes, a source code was
written, implemented and presented in the paper. The source code was intensively tested while
performing a simulation of a mobile robot localization problem in a manufacturing environment
and it prooved its usefulness. However, the experimental setup built on LEGO Mindstorms™
NXT technology, being prepared at the moment, should test the proposed approach and revel

Robots in Manufacturing Environments 493


its advantages and disadvantages. Having previous analysis in mind, proper implementation of
high level robot arhitecture /1,12,14/ consisting of individual software modules is being
considered and analyzed in order to develop an architecture capable of solving localization,
path planning, obstacle avoidance and perceptual interpretation as basic building blocks of a
mobile robot operating on the shop floor as an inherent dynamic environment.

ACKNOWLEDGEMENTS

This paper is a part of the research called: Flexible Automation and Implementation of Intelligent
Manufacturing Systems in Sheet Metal Production, being financed by The Government of the
Republic of Serbia, The Ministry of Science (2008 - 2010). The authors would like to thank their
partners in SAGA Company – Belgrade as well for help and much needed support.

7. REFERENCES

1. Arkin R., Behavior-Based Robotics, MIT Press, Cambridge, Massachusetts, 1998.


2. Babić B., FLEXY – Intelligent System for Design of FMS (in Serbian), The series of scientific
monograph books Intelligent Manufacturing Systems (Editor: Prof. Dr Vladimir R. Milačić),
Vol. 5 (ISBN 86-7083-250-X), University of Belgrade-Faculty of Mechanical Engineering,
1994.
3. Bar-Shalom Y., Li X.R., Kirubarajan T., Estimation with Applications to Tracking and
Navigation-Theory, Algorithms and Software, John Wiley & Sons, 2001.
4. Borenstein J., Everett H.R., Feng L., Where am I? Sensors and Methods for Mobile Robot
Positioning, Ann Arbor, University of Michigan. Available at http://www-
personal.engin.unimch.edu/~johannb/position.htm
5. Choset H., Lynch K.M., Hutchinson S., Kantor G., Burgard W., Kavraki L.E., Thrun S.,
Principles of Robot Motion-Theory, Algorithms and Implementations, MIT Press, Cambridge,
Massachusetts, 2005.
6. Craig J.J., Introduction to Robotics, Mechanics and Control, Addison-Wesley Publishing
Company, 1989.
7. Fukuda T., Arakawa T., Intelligent Systems: Robotics versus Mechatronics, Annual Reviews
in Control, 22, pp. 13-22, 1998.
8. Groover, M.P., Automation, Production Systems, and Computer-Integrated Manufacturing,
2nd Edition, Prentice Hall, 2001.
9. Kalman R.E., A New Approach to Linear Filtering and Prediction Problems, Transactions of
the ASME-Journal of Basic Engineering 82 (Series D), pp. 35-45., 1960.
10. Kopacek P., Intelligent Manufacturing: Present State and Future Trends, Journal of
Intelligent and Robotic Systems, Vol.26, pp. 217-229,1999.
11. Miljković Z., Systems of Artificial Neural Networks in Production Technologies (in Serbian),
The series of scientific monograph books Intelligent Manufacturing Systems (Editor: Prof. Dr
Vladimir R. Milačić), Vol. 8 (ISBN 86-7083-455-3), University of Belgrade-Faculty of
Mechanical Engineering, 2003.
12. Murphy R.R., Introduction to AI Robotics, MIT Press, Cambridge, Massachusetts, 2000.
13. Russell S.J., Norvig P., Artificial Intelligence: A Modern Approach (second edition), Prentice
Hall, Englewood Cliffs, NJ, 2002.
14. Siegwart R., Nourbakhsh I.R., Introduction to Autonomous Mobile Robots, MIT Press,
Cambridge, Massachusetts, 2004.
15. Thrun S., Burgard W., Fox D., Probabilistic Robotics, MIT Press, Cambridge,
Massachusetts, 2005.

494 3rd ICMEN – 2008

View publication stats

You might also like