Optical Guidance System for Multiple Mobile Robots
Igor E. Paromtchik, Hajime Asama
The Institute of Physical and Chemical Research (RIKEN), Advanced Engineering Center
Hirosawa 2-1, Wako-shi, Saitama 351-0198, Japan
This paper describes our research work towards the de-
velopment of an optical guidance system for multiple mo-
bile robots in an indoor environment. The guidance sys-
tem operates with an environmental model, communicates
with mobile robots and indicates their target positions by
means of a light projection from a laser pointer onto the
ground. Processing the image datafrom a CCD color cam-
light beaconon the ground and estimate its relative coordi-
nates. The robot’s control system ensures the accurate mo-
tion of the robot to the indicated target position. The guid-
ance system subsequently indicates target positions corre-
sponding to a desired route for a specified mobile robot in
the fleet. The concept of the optical guidance system, its
implementation and experimental results obtained are dis-
Guidance of a mobile robot involves its localization in
the environment . The precise localization becomes es-
pecially relevant in the case of multiple mobile robots shar-
ing a common environment. Various localization methods
are known, from simple and widely used odometry and
other dead reckoning methods to active and passive range
sensing approaches;see a recentsurveyon laser rangefind-
ers, triangulation range finders and passive stereo for mo-
bile robots .
vision) along with external means (landmarks, beacons)
and fusion of sensor data are necessary in order to obtain
the precise position and orientation of the robot in the en-
vironment and update the environmental model. An in-
crease in discrepancybetween the actual robot positionand
its estimate (e.g. if localization relies on a dead-reckoning
method) can lead to inadequate motion planning and con-
trol, resulting in collisions with objects or other robots.
In order to deal with the localization problem, various
optical guidance methods have been developed such as us-
ing reflective beacons, tracking stationary light sources,
tracking a guidance line on the floor or ceiling, or using
a scanning laser on the mobile robot in order to measure
distances to surrounding objects .
We propose an optical guidance system for mobile
robots that makes use of a projected laser light. The guid-
ance system operates with the environmental model and
comprises a computer-controlledlaser pointer with at least
two degrees-of-freedomin orderto direct a laser beamonto
the desired positions on the ground. The guidance system
communicates with the mobile robot when indicating its
target position and subsequent checking if the robot has at-
tained this position.
The key idea of the optical guidance system is to indi-
cate the numerical coordinates of the target position for the
mobile robot by means of projection of a laser light onto
the ground. The on-board vision system of the robot pro-
cesses the color images in order to detect the laser light
beacon on the ground and evaluate its relative coordinates.
This visual feedback ensures the accurate following of the
indicated positions by the robot.
The main advantage of the proposed guidance system is
the improved accuracy. The system also allows implicit
localization of the mobile robot within the environment:
when the robot has reached its indicated target position, an
estimate of its coordinates in the environmental model is
known. Since the robot’s control system operates with the
relative coordinates of target positions obtained from im-
age processing, the transformation between the coordinate
systems of the environmental model (“world” coordinate
system) and that of the mobile robot becomes less relevant
Our paper focuses on the concept of the optical guid-
ance system integrated into an environment with multiple
mobile robots. The paper is organizedas follows. The con-
cept of the proposed optical guidance system is described
in section 2. The mobile robot and its control architecture
are presented in section 3. The operation of the guidance
system is discussed in section 4. The implementation and
our experimental results are presented in section 5. The
conclusions are given in section 6.
Proceedings of the 2001 IEEE
International Conference on Robotics & Automation
Seoul, Korea • May 21-26, 2001
0-7803-6475-9/01/$10.00© 2001 IEEE
2The Optical Guidance System
The proposed optical guidance system indicates target
positions for the mobile robotby means of a laser light pro-
jected onto the ground. The guidance system is sketched in
Fig. 1. The system comprises a teleoperation board con-
nected to a laser pointer which has at least two degrees-
of-freedom in order to direct the optical axis of the laser
to any position on the ground. The coordinates of the tar-
get positions are computed from the environmental model
accordingto the motiontask, or theyare set bya humanop-
erator from the teleoperation board. The guidance system
relies on a TCP/IP client-server wireless communication
with the control systems of the mobile robots. The robot’s
vision system processes color images in order to detect the
laser light beacon on the ground and evaluate its relative
Mobile Robot i
Mobile Robot 1
Mobile Robot N
Figure 1: A sketch of the optical guidance system
? denote a common coordinate system in
Fig. 1, where
for a mobile robot, and
pointer. Let the laser pointer be situated at a hight h0from
the ground and have two degrees-of-freedom
light beam from the laser pointer will indicate the coordi-
laser pointer is obtained from the following expressions:
? are coordinates of a target position
? are coordinates of the laser
? on the ground, if an orientation
? of the
The featuresof the proposedoptical guidancesystem are
summarized as follows:
? target positions can be indicated precisely in the en-
vironment by means of a laser pointer connected to a
? close-loop control based on visual feedback provides
better positioning accuracy of the mobile robot.
? the path to follow can be indicated as a sequence of
? TCP/IP and wireless Ethernet are used for commu-
nication based on a client-server model between the
guidance system and the mobile robots.
? accumulation of positioning errors will not influence
cause the localization is performedwhen the robothas
attained its indicated target position.
? one guidance system can indicate target positions for
multiple mobile robots in the environment.
The communication ability and updating the environ-
mental model in the guidance system allow us to use this
system as a mediator for multiple mobile robots. For in-
stance, the sensor data gathered by the robots and stored
in the environmental model is available to all robots in the
fleet, i.e. cooperative knowledge acquisition and sharing is
achieved. The distribution of the motion tasks and their al-
location to the mobile robots are performed with the use of
the environmental model as a part of the guidance system.
One mobilerobotis also able to requestthe system to guide
another robot to a specified destination.
To the best of our knowledge, such an optical guidance
system is new and it presents an alternative approach to
the robot guidance. This system is especially intended for
an indoorenvironmentwherethe globalpositioningsystem
(GPS) can not be used.
3The Robot’s Control Architecture
The overall control architecture of the developed mobile
robot which is capable of operating with the optical guid-
ance system, is shown in Fig. 2. It involves three main
parts: the vehicle with its actuators and sensors, the on-
board real-time control system, and the remote control in-
The mobile robot is equipped with four omnidirectional
wheels which allow it to perform motions in two direc-
tions and rotate simultaneously. Three DC motors and a
transmission mechanism provide the omnidirectional mo-
tion . Three servo-systems execute the motion control
commands issued by the control system of the robot and
ensure attaining the commanded position, orientation and
The sensor system of the robot involves a CCD color
camera and eight infrared and sonar range sensors. These
sensors gather data about the local environment in order to
evaluate relative distances between the robot’s frame and
environmental objects. Processing of the color data about
Actuators and Sensors
Figure 2: The overall control architecture
the local environment allows the robot to detect and local-
ize a specified object, e.g. another mobile robot or a pro-
jected laser light. The sensor system also involves three
encoders used in the feedback loop of the servo-systems.
The control system processes the data obtained from the
sensor system and the servo-systems and issues the motion
control commands which are subsequently executed by the
servo-systems. The robot is able to follow a specified ob-
the object as a function of the motion velocity. The target
positions for the mobile robot can also be set by a remote
computeror an operatorfroma graphicaluser interface.
The reactive control ensures an autonomous operation
of the robot in a dynamic environment .
sion avoidance algorithm operates with a rule matrix ob-
tained froman adaptivebehavioracquisitionscheme which
is based on reinforcement learning. The reactive control
processes data gathered by the ultrasonic and infrared sen-
sors as well as the CCD color camera. Based on the sen-
sors configuration, eight possible directions of motion are
considered [5, 6]. The reactive control algorithm ensures a
collision-free motion to a given target position.
The motion controller provides coordinated translation
androtationof the omnidirectionalvehiclewhenmovingto
a target position, as well as to attain a desired velocity .
Omnidirectional motion (holonomic case) or constrained
motion (non-holonomic case) can be performed according
to the assigned task. For instance, vision-based tracking of
a dynamic object requires the object to be kept in the view
field of the on-boardcamera and is achieved by the coordi-
nated control of the translational and rotational coordinates
of the mobile robot.
The performance monitoring aims to increase the relia-
bility andsafetyof the robotoperation. Monitoringofmea-
surable signals allows for fault detection and diagnosis in a
closed-loopduringoperation , e.g. measuringthe capac-
ity level of the electric batteries of the robot and, if needed,
requesting the control system to interrupt execution of the
on-going task and direct the robot to a location where the
electric batteries can be replaced.
4 The Robot Guidance
The optical guidance system proposed in section 2 in-
dicates the target position as a laser light beacon on the
ground. The robot’s control system detects this beacon
from image processing and generates a smooth trajectory
can also be set numerically in teleoperation mode). The
guidance system operates according to the following basic
algorithm in order to indicate a target position for a mobile
? to the detected target position xd(the target position
1. Establish a client-server connection between the tele-
operation board and control system of the mobile
2. Transmit a request whether the robot’s control system
is ready to process a new target position.
3. If the mobile robot is ready to receive a new target
position, then set the laser pointer in the appropriate
?. Otherwise, go to step 2.
4. Turn on the laser pointer light in order to indicate the
target position on the ground.
5. If the robot’s control system confirms detection of the
indicated position, then turn off the laser light. Other-
wise, wait until the confirmation or failure response is
received from the robot’s control system.
6. If the indicated target position could not be detected
(failure response),then proceedto failure analysis and
its compensation, e.g. by means of setting a target po-
sition closer to the mobile robot.
7. Ifanothertargetpositionmust beset, thengotostep 2,
The light of a laser pointer is specified by its brightness
(power output), wavelength (color) and focus. The laser
light projected on the ground is seen as a small, bright dot
of red light (we use a laser with a wavelength of 635 nm
and a power output of 2 mW). The size of such a beacon
is within the known range of a given laser (e.g. a spot of
8-10 mm in diameter at a 10 m distance). The detection of
the beaconcan be performedeither by means of comparing
two images obtained when the laser is “off” and “on” ,
or by means of a search for an area of specified brightness,
color and size in the captured image.
Since the brightnessat the laser beaconchangesabruptly
in magnitude, an edge detection technique is applied .
We have tested a discrete differencing (horizontal and ver-
tical) of the RGB image and a subsequent search for pix-
els where the intensity change of the red color is maximal.
When such pixels are found, an image segmentation by
means of a global thresholding technique is performed: the
neighboringpixels in the original image are evaluated rela-
tive to a given color threshold in order to estimate the size
of detected areas. The selection of pixels which correspond
to the red laser light makes use of a CIE chromaticity dia-
gram where the chromaticity values of the given laser are
within a known range. The selected area is centered, and
the coordinate transformation from pixels into meters re-
sults in obtaining the target position xd.
The subsequent generation of the smooth trajectory x?t
to the target position xdis based on our motion generation
approach . This approach represents a modified cubic
spline-interpolation which involves a recomputation of the
spline betweenthe actualpositionof the mobilerobotanda
virtualrunningpointsituatedat a distance∆xTaheadof the
robot in the target direction xd, as it is illustrated by Fig. 3.
Figure 3: A virtual running point
The virtual point is shifted with a sampling period T
in the direction xd in order to lead the robot to the tar-
get. The recomputationof the spline terms in the proximity
∆xsof xd. This motion generation approach is effectively
used for various tasks where trajectory generation in real
time is needed to control the mobile robots, e.g. collision
avoidance, tracking an object or a laser light beacon on the
ground as it is performed in the optical guidance system.
The mobile robot is equipped with a CCD color Toshiba
camera (focal length is 7.5 mm) that acquires images about
the local environment. At present, we use a LP-310 Plus
laser pointer(wavelengthis 635nm, poweroutputis 2mW)
that provides a red light. In order to direct the laser beam
onto the the desired positions on the floor, this laser is
mountedonto a pan-tilt mechanismequippedwith two step
motors of a Canon communication camera VC-C1, as it is
shown in Fig. 4.
Figure 4: A laser pointer mounted onto a camera with a
The experimental setup is sketched in Fig. 5, where the
robot’s camera is in the inclined position relative to the
ground. The laser light beacon is seen in the image as
a bright spot of the size of a few pixels, as illustrated in
Fig. 6 and Fig. 7. The projected laser light on the ground
has a blur contour and the shape of the beacon is elliptical,
shown in Fig. 7. The central part of the beacon in the im-
age is white-colored that shows saturation of the camera.
The pixels indicated by bold contours in Fig. 7 illustrate
the horizontal and vertical differencing applied in order to
find the pixels where the intensity change of the red color
One should note that detection of a laser light beacon
depends strongly on the lighting conditions and the sur-
face material. For instance, our experiments on a grey con-
Figure 5: A sketch of the experimental setup
Figure 6: An example of a camera snapshot of the ground:
a laser light beacon(center),a 5 Yen coin (left)and a 30 cm
crete surface or a green synthetic carpet have shown reli-
able detection. The maximal detection distance depends
on the laser power output and the camera sensitivity. The
precision of the position estimation of the detected laser
light beacon is influenced by various factors such as quan-
tization errors due to the small size of the beacon, camera
displacement from the calibrated setting as well as camera
con from pixels into meters, a non-linear transformation
Figure 7: A laser light beacon (zoomed)
table was experimentally obtained for a rectangular area in
front of the robot’s camera. The measurement results for
an area of 1.0 m in width and 1.4 m in length (the size of
the graph paper used) are depicted in Fig. 8. The curves of
the longitudinal distance are plotted for the lateral distance
rangefrom -0.4m to 0.4 m. The functionsof the lateral dis-
tance are plotted for the longitudinal distance range from
0.25 m (the shortest line in Fig. 8,b) to 1.3 m.
Longitudinal distance [m]
Longitudinal distance [pixel]
Lateral distance [m]
Lateral distance [pixel]
Figure 8: Coordinate transformation from pixel to meter:
a - longitudinal distance, b - lateral distance
The estimation of the coordinates of the laser light bea-
con on the ground relative to the robot’s camera is illus-
trated in Fig. 9. The camera position corresponds to the
originof the coordinatesystem in this figure, and the actual
positions where the laser light is projected are depicted as
points. The plus signs in Fig. 9 show the correspondingco-
ordinates estimated by image processing. The localization
accuracy is lower at the borders of the camera view field
(upper and lower point sequences in Fig. 9) while the ac-
curacy along the optical axis of the camera (central points
in Fig. 9) is sufficient for tracking the projected laser light
(the maximal dispersion was less than 10 mm, and the av-
erage dispersion was about 3 mm).
0.2 0.40.60.81 1.2 1.4
Lateral distance [m]
Longitudinal distance [m]
Figure 9: Localization of projected laser light
The software for the teleoperation and optical guidance
system is developedin JAVA language. The software of the
robot’s control system is implemented in C language and
runs under VxWorks real-time operating system on a Pen-
tium 200 MHz processor. The client-server communica-
tion between the teleoperation and optical guidancesystem
(“client”) and the mobile robots (“servers”) is performed
via a wireless Ethernet. Our video illustrates the experi-
The concept of the optical guidance system that makes
use of a laser pointer was introduced. The features of the
optical guidance system and its use for multiple mobile
robots were discussed. The control architecture of the mo-
ered. The implementation and experimental results on the
operation of the optical guidance system were presented.
Our future work will deal with the improvement of the im-
age processing algorithm, system integration and conduct-
ing experiments with multiple mobile robots.
 J. Borenstein, H. R. Everett, L. Feng, Navigating Mo-
bile Robots – Systems and Techniques, A K Peters,
Wellesley, MA, USA, 1996, 225 p.
 M. Hebert, Active and Passive Range Sensing
for Robotics, Proc. of the IEEE Int. Conf. on
Robotics and Automation, San Francisco, CA, USA,
April 24-28, 2000, pp. 102-110.
 H. Asama et al.,
Directional Mobile Robot with 3 DOF Decou-
pling Drive Mechanism, Proc. of the IEEE Int.
Conf. on Robotics and Automation, Nagoya, Japan,
May 21-27, 1995, pp. 1925-1930.
Development of an Omni-
 T. Suzuki et al. A Multi-Robot Teleoperation System
Utilizing the Internet, Advanced Robotics, Vol. 11,
No. 8, 1998, pp. 781-797.
 Y. Araiet al., AdaptiveBehaviourAcquisitionofCol-
lision Avoidance among Multiple Autonomous Mo-
bile Robots, Proc. of the IEEE/RSJ Int. Conf. on
Intelligent Robots and Systems, Grenoble, France,
Sept. 7-11, 1997, pp. 1762-1767.
 I.E.Paromtchik,U.M.Nassal, ReactiveMotionCon-
trol for an Omnidirectional Mobile Robot, Proc. of
the Third European Control Conference, Roma, Italy,
Sept. 5-8, 1995, pp. 3074-3079.
 I. E. Paromtchik, H. Asama, A Motion Genera-
tion Approach for an Omnidirectional Vehicle, Proc.
of the IEEE Int. Conf. on Robotics and Automa-
tion, San Francisco, CA, USA, April 24-28, 2000,
 S. Okina et al., Self-Diagnosis System of an Au-
tonomous Mobile Robot Using Sensory Information,
 A. Nakamura, T. Arai, T. Hiroki, J. Ota, Deve-
lopment of a Multiple Mobile Robot System Con-
trolledbya Human–Realisation of ObjectCommand
Level, Distributed Autonomous Robotics Systems 4,
L. E. Parker, G. Bekey Eds., Springer Verlag Tokyo,
2000, pp. 209-218.
 G. X. Ritter, J. N. Wilson, Handbookof Computer Vi-
sion Algorithms in Image Algebra, CRC Press, Boca
Raton, USA, 1996.
 C. C. Yang, F. W. Ciarallo, Optimized Sensor Place-
ment for Active Visual Perception, J. of Robotic Sys-
tems, Vol. 18, 2001, pp. 1-15.