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

net/publication/223943640

Developing a mobile robot for transport applications in the hospital domain

Article in Robotics and Autonomous Systems · July 2010

DOI: 10.1016/j.robot.2010.03.010 · Source: DBLP

CITATIONS READS

71 627

5 authors, including:

Masaki Takahashi Toshiki Moriguchi

Keio University Murata Machinery Ltd., Japan, Kyoto
415 PUBLICATIONS 1,448 CITATIONS 18 PUBLICATIONS 162 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Masaki Takahashi on 11 November 2020.

The user has requested enhancement of the downloaded file.

 Robotics and Autonomous Systems 58 (2010) 889–899

Contents lists available at ScienceDirect

Robotics and Autonomous Systems

journal homepage: www.elsevier.com/locate/robot

Developing a mobile robot for transport applications in the hospital domain

Masaki Takahashi a,∗ , Takafumi Suzuki a , Hideo Shitamoto b , Toshiki Moriguchi b , Kazuo Yoshida a,1
a
Department of System Design Engineering, Keio University, Yokohama, 223-8522, Japan
b
Muratec R&D Center, Murata Machinery, Ltd., Kyoto, 612-8686, Japan

article info abstract

Article history: We have been developing MKR (Muratec Keio Robot), an autonomous omni-directional mobile transfer
Available online 27 March 2010 robot system for hospital applications. This robot has a wagon truck to transfer luggage, important
specimens, and other materials. This study proposes an obstacle collision avoidance technique for the
Keywords:
wagon truck pulling robot which uses an omni-directional wheel system as a safe movement technology.
Motion planning
Moreover, this paper proposes a method to reach the goal along a global path computed by path planning
Path tracking
Obstacle avoidance without colliding with static and dynamic obstacles. The method is based on virtual potential fields.
Virtual potential field method Several modules with different prediction times are processed in parallel to change the robot response
Prediction time according to its relative velocity and position with respect to the obstacles. The virtual force calculated
Omni-directional mobile robot from each potential field is used to generate the velocity command. Some experiments were carried out
to verify the performance of the proposed method. From the experimental results in a hospital it was
confirmed that the robot can move along its global path, and reach the goal without colliding with static
and moving obstacles.
Published by Elsevier B.V.

1. Introduction Various essential technologies, such as a self-localization

scheme, an environmental map formation and path planning, a
There is a common thought that the new robotic market ex- learning algorithm and communication, have been developed for
tends the use of robots from industrial applications for manufac- autonomous mobile robots used in the service robotics field.
turing processes to a variety of applications that use moving robots A variety of service robots which provide service in the actual
in an open environment where persons can be present. Moreover, environment with other moving objects, including people, have
if new advanced technologies are being developed because there been developed [1–5]. It has been reported that MINERVA realizes
are many social needs for transfer robots, the practical use of robots a tour guide through a museum without colliding with people [1].
has the possibility of becoming a major breakthrough. The technol- There were several early works on path planning with moving
ogy concerning safety is important, though there are various tech- obstacles [6,7]. Several typical robot navigation and collision
nical problems to overcome.
avoidance methods have been proposed [8–12].
In our research project, we have developed a technology suit-
A variety of tasks are required for service robots, but here
able for a robot that transports surely and safely luggage, important
we will focus on problems related to moving, which is the most
specimens, and other materials, in the public facilities of a hospital.
fundamental and important task. If the environment includes
This robot, shown in Fig. 1, is called MKR (Muratec Keio Robot). To
ensure mechanical safety, the robot has an omni-directional wheel humans, both safe and efficient movement is required. This study
system as shown in Fig. 2. In addition, the robot takes into account proposes an obstacle collision avoidance technique for a robot
safety by using dynamic obstacle avoidance and active safety tech- that pulls a wagon truck using an omni-directional wheel system
nology since it has to operate in an open environment. We have as a safe movement technology. When the robot moves in a
adopted as a transportation method pulling a wagon truck, which dynamic environment, the time spent to deal with obstacles and
can correspond to various transportation capabilities, in addition the required rapidity of response are different according to the
to transportation with the robot itself, as shown in Fig. 3. situation.
To achieve a smooth obstacle avoidance ensuring safety in the
presence of moving obstacles including humans, we propose a hi-
∗ Corresponding author. Tel.: +81 45 566 1660; fax: +81 45 566 1660. erarchical action control technique based on the prediction time
E-mail address: [email protected] (M. Takahashi). of the action. Fig. 4 shows the concept of the proposed technique.
URL: http://www.yt.sd.keio.ac.jp/ (M. Takahashi). The hierarchical control method considers various time scales for
1 Prof. Kazuo Yoshida died in 2008. actions such as goal path planning, obstacle avoidance within the
0921-8890/$ – see front matter. Published by Elsevier B.V.
doi:10.1016/j.robot.2010.03.010
890 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899

(a) MKR-001. (b) MKR-002. (c) MKR-003.

Fig. 1. Hospital autonomous omni-directional mobile transfer robot system: MKR (Muratec Keio Robot).

Fig. 2. Robot and sensor.

recognizable range, and emergency avoidance to deal with unex- As shown in Fig. 5(a), there is a danger of colliding with static
pected events. In the proposed method, several modules for each and dynamic obstacles when the robot is pulling the wagon truck
action are composed in parallel. The vertical axis is the time scale during the collision avoidance operations to change the direction
in the control system. In the lowest module, the robot can move to of movement along the motion path. To achieve a movement that
the goal safely and efficiently by planning from the environment considers safety and that is appropriate for the narrow aisle shown
information which is obtained in advance. In the higher module,
in Fig. 5(b), we propose a method which uses the features of an
the robot moves more safely by using the estimated information
omni-directional movement mechanism. In addition, we propose
about the obstacles to avoid them. By integrating the output of
each module comprehensively, it is possible to realize a safe and a method to reach the goal along a global path computed by path
efficient movement according to the situation. In this paper, as an planning while avoiding the obstacles.
example of the proposed moving control technique, we present a To verify the effectiveness of the proposed technique, some
method based on virtual potential fields [13–15]. experiments using the real apparatus of an autonomous omni-
 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899 891

Table 1
Specification of hospital autonomous omni-directional mobile transfer robot
system (MKR-003).

Width 550 mm
Length 750 mm
Height 1250 mm
Weight 70 kg
Maximum Velocity 1.0 m/s
Driving system Omni-directional wheel with four-wheel
drive and independence suspending
suspension
Arm 3 DOF
Head 2 DOF
Sensor 1 Laser range finder
16 Ultrasonic sensors
1 Stereo camera
Bumper switch
Continuous operating time 1h
Charging time 1.5 h
Fig. 3. Traction type autonomous omni-directional mobile robots.
2. Hospital autonomous omni-directional mobile transfer robot
system

2.1. Concept

We have been developing the MKR (Muratec Keio Robot)

as a hospital autonomous omni-directional mobile transfer
robot system since 2005. To deal with various transportation
requirements in the hospital, this robot adopts a truck pulling type
as shown in Fig. 3.
Technological developments that allow the robot to move
safely in the environment where there are some persons is an
important required technology. This requirement is important not
only with regard to the robot movements, but also for human-
robot interaction, such as recognizing human gestures, since the
robot is designed to operate in an open hospital environment. In
the future we will develop additional movement technology to
Fig. 4. Problem for an autonomous mobile robot. increase safety from the robot side by using human gestures in
addition to the method we present here.

2.2. Hardware system

The hospital autonomous omni-directional mobile transfer

robot system is shown in Fig. 2. The robot has an omni-drive system
with four roller wheels, shown in Fig. 2. The four-wheel omni-drive
system is utilized to increase the freedom of the moving direction
and to improve the stability of the robot while it is moving. It has
a Laser Range Finder (LRF), sixteen ultrasonic sensors, and a stereo
vision system as external sensors (Table 1).

2.3. Operation scenario

The MKR (specifically the MKR-003) has functions like mapping,

localization, path planning, and collision avoidance. Its operation
scenario is as follows:
1. The operator puts luggage in the wagon and inputs the goal
point.
2. The robot plans the path that connects the starting point with
the goal point.
3. The robot transports the luggage to the goal point along the path
computed by the path planning algorithm. While transporting,
the robot avoids any unexpected obstacles and then returns to
the planned path.
Fig. 5. Collision avoidance of an autonomous robot. 4. When the robot reaches the goal point, another operator takes
the luggage out of the wagon.
directional robot have been carried out. Moreover, experimental In this paper, we focus on the collision avoidance algorithm and
results in a hospital environment are shown. path tracking.
892 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899

study, α is 45 degrees. The velocity command in the translational

dφ direction is decided depending on the proposed hierarchical action
control technique. To prevent interference with the environment
c0
dx
c1 and to achieve safe transportation, the desired angle dφ , shown
S in Fig. 6, is decided. The robot calculates a shortest point to the
planned global path, and the tangent vector at this point. As shown
θ in Fig. 7, the desired angle dφ is the angle between the direction of
the tangent vector and the present direction of the robot. When
α = 45 deg the robot avoids the obstacle, the robot moves to keep the desired
angle as shown in Fig. 5.
dy
L
3.3. Design concept

c3 We propose a hierarchical moving control method that consid-

c2
ers differences in time scale among actions to achieve both safe
and effective movement; e.g., the movement of the robot over the
shortest distance can be realized based on predicting the move-
Fig. 6. Platform of omni-directional mobile robot. ments of the objects, such as those that have a relatively large time
scale, including structures in the environment that do not change,
3. Hierarchical action control technique or moving obstacles that only move in a constant and predictable
direction. However, in a dynamic environment with moving ob-
3.1. Problems to be solved jects such as humans, spontaneous events may occur that cannot
be predicted, and thus it is desirable to use very short-term predic-
In this study, we address the issues concerning robots that tions to realize a safe movement. To realize a movement over such
are expected to move within general buildings, such as public short distances safely, decision making regarding actions which
spaces, offices, hospitals, or homes, with other moving objects, considers the time scale differences is required. The hierarchical
including people. A variety of tasks are required for such service control method we propose considers various time scales for ac-
robots, but here we will focus on problems related to moving, tions in cases with different time scales, such as goal path planning,
which is the most fundamental and important of the tasks. As obstacle avoidance within the recognizable range, and emergency
service robots come in direct contact with humans, we must avoidance to avoid spontaneous events. The module is a potential
attach importance to safety to ensure that the robots will not function with the time scale as a design parameter, and generates
harm humans, while minimizing as much as possible the time an artificial potential field.
required for the movements. It is desirable for the robot to reach its
goal without colliding with obstacles, including humans. However,
3.4. Design method
this type of environment includes events that are unpredictable
at the design phase, and it is difficult to respond to various
In this study, the potential function was designed based on the
circumstances using a scenario-based method. Thus, we propose
a moving control method that generates actions corresponding to repulsive potential reported by Khatib [13]:
the circumstances, by embedding a fundamental problem-solving U = Uxg + Uo (5)
method for moving into the control system.
where U is an artificial potential field. xg is the goal position, Uxg is
3.2. Omni-directional wheel an attractive potential field, and Uo is a repulsive potential field.
In the proposed method, the repulsive potential field is defined
The platform of the robot with omni-directional wheel system with respect to the time scale Ti . Several modules that generate the
is shown in Fig. 6. The relationship between the distance moved repulsive potential field are structured hierarchically.
by the robot (dx, dy) and the distance moved by each wheel ci (i =
0, 1, 2, 3) is as follows: 3.4.1. Path tracking to the global path of path planning
dx = S cos θ

In this section, we describe the method to reach the goal along
(1)
dy = S sin θ a global path. To achieve both path tracking to the global path
and obstacle collision avoidance, some subgoals are set on the
cos α sin α
   
c0 L global path. In this method, the attractive potential field which
dx
" #
c1  cos α − sin α −L  is generated by each module is different. The robot generates an
c  = − cos α dy (2)
− sin α L 
attractive potential to the subgoal, and reaches the goal while
2
dφ
c3 − cos α sin α −L updating the subgoal according to the situation.
 1 1 1 1 Uxsg = ka ρg

(6)
 cos α cos α cos α cos α  c0
 
dx
" #
1 1

1 1 1  c1 
 where xsg is the subgoal position and Uxsg is the attractive potential
dy =  − − (3) field. In this study, we decided to use as a subgoal the point where
4  sin α sin α sin α sin α  c2
 
dφ  1  c3 the circle of two meters of radius that is centered on the robot
1 1 1
− − intersected the global path, as shown in Fig. 7.
L L L L
p
S= dx2 + dy2 (4) 3.4.2. Collision avoidance to obstacles
where L is the distance between the center of gravity of the robot To achieve a safe movement the robot should predict the
and the wheel, and α is the inclination angle of the wheel. In this positions of the obstacles. To achieve a safer movement, the
 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899 893

y In the proposed method, the repulsive potential field Uo is

defined with respect to the time scale Ti .
Goal point ρTi is the predicted shortest distance between the robot and
Global path of
the obstacle Oj at time scale Ti . This is determined by calculating
path planning
the predicted positions of the robot and the obstacle after time Ti
using the current velocity, and by calculating the shortest distance
in these positional relationships.
Robot dφ
1/ Ti vr ,oj + ρ0 in Eq. (8) is a parameter that determines the


Subgoal
range of the potential field in the repulsive potential of Khatib. ρ0 is
the limit distance of the potential field influence and η is a constant
gain. By considering the relative velocity, the response of the robot
is expected to change even if the relative position of the obstacle
Start point
to the robot is the same. For example, when the relative velocity is
x
large, the robot reacts more quickly.
Fig. 7. Subgoal on the global path produced by the path planning method. The exponential part 1/Ti is the part of the equation that
determines the priority level of the output of each module. When
the robot faces an emergency situation, the time scale is small, and
prediction time should change according to the relative position
the output of the module with a small time scale is large. Thereby,
and velocity between the robot and the obstacles. It is not easy
the robot acts more rapidly.
to estimate a suitable prediction time according to the situation.
Therefore, in the proposed method, several modules with different
prediction times are processed in parallel. 3.4.3. Integration
The response and action of the robot should change according to A force for the point x of the robot is derived from the following
its relative velocity even if the relative position of the obstacle with equation:
respect to the robot is the same. For example, when the relative ∂U
velocity is large, the robot should react from an earlier stage, and F(x) = − (9)
act more quickly. Therefore in the proposed method, the selection ∂x
condition of the module is changed based on the relative velocity where ∂∂Ux denotes the partial derivative vector of the total virtual
and prediction time. potential U. From Eqs. (6) and (9), the attractive force Fxg which
The module is a potential function with the time scale as a allows the position x of the robot to reach the goal position xg is as
design parameter, and generates a potential field. Specifically, the follows:
potential function was designed based on the repulsive potential
∂ρg
reported by Khatib [13]. Fxg = −ka . (10)
X ∂x
Uo = UoTji (7) T
From Eqs. (8) and (9), the repulsive force Foij to the obstacle Oj is as
 ! T1 follows:
i
 1 1 ! T1 −1
η T
 

i − i
1 ∂ρTi
ρTi Ti vr ,oj + ρ0 1 1

UoTji = (8) η

 −


 if ρTi ≤ Ti vr ,oj + ρ0 FToij = ρTi Ti vr ,oj + ρ0 ρT2i ∂ x (11)
if ρTi > Ti vr ,oj + ρ0 if ρTi ≤ Ti vr ,oj + ρ0

0



if ρTi > Ti vr ,oj + ρ0 .

0

Fig. 8. Output of each module and integration.

894 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899

(a) Trajectory of the robot. (b) Cutoff situations of the collision avoidance.

(c) Time history of the module activations.

Fig. 9. Experimental results in the case of static obstacles.

When there are several obstacles in the environment, the We assume that the robot is a point mass, and the virtual forces
repulsive force Fo is sum of the repulsive forces to each obstacle. derived from several potential fields generated by each module are
X combined. The velocity vector is determined based on the force.
Fo = FToij . (12) The potential approach has a vibration problem caused by the
magnitude of velocity and roughness of the control period. Thus,
The command vector F of the robot is derived from the following a low pass filter on each element of the virtual force output in each
equation: module is used to suppress such vibration as shown in Fig. 8(a). In
this study, we used the filter that uses the reciprocal of each time
F = Fxg + Fo . (13) scale as a cut-off frequency.
 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899 895

(a) Trajectory of the robot. (b) Cutoff situations of the collision avoidance.

(c) Time history of the module activations.

Fig. 10. Experimental results in the case of moving obstacles.

3.4.4. Decision method of the time scale 0.1 s to 5.0 s at intervals of 0.1 s. From the simulation study, we
In this study, we designed two modules to generate the repul- decided to use T1 = 1.2 s and T2 = 0.5 s.
sive potential field. We need to set the time scale of each module,
T1 and T2 . Since it is not easy to decide how to set the values of
4. Experiments
these parameters systematically, we decided them by a simulation
study. In the parameter study, the robot moves to the goal in an
environment which includes static and dynamic obstacles. In the To verify the performance of the proposed collision avoidance
simulations, the robot velocities are 0.5, 0.75, 1.3 m/s, and the ob- technique to static and dynamic obstacles in various situations,
stacle velocities are 0.0, 0.50, 0.75, 1.3, 2.5 m/s. T is changed from experiments using the real robot were carried out. In order to
896 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899

(a) Trajectory of the robot. (b) Cutoff situations of the collision avoidance.

(c) Time history of the module activations.

Fig. 11. Experimental results in the case of a person’s sudden appearance.

recognize the environment, the robot has external sensors, such 4.1. Case 1: static obstacles
as a stereo camera, laser range finder and ultrasonic sensors.
However, in this research the robot recognizes the environment Fig. 9 shows the experimental environment to verify the
making use only of the laser range finder. The velocity limit of the effectiveness of the proposed method with two static obstacles.
robot is 0.5 m/s and the acceleration limit is 1.0 m/s2 . The initial position of the robot is (0 m, 0 m). The obstacle size is
Moreover, to verify the performance of the path tracking and 0.20 m × 0.33 m × 0.50 m and their initial positions are (−0.5 m,
collision avoidance in the hospital, we carried out some experi- 1.5 m) and (0.5 m, 3.0 m) respectively. Fig. 9(a) shows the trajectory
ments in the hospital. of the robot. Fig. 9(b) shows the cutoff situations of the collision
 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899 897

(a) Map. (b) Robot path.

Fig. 12. Map of the floor in the hospital and path produced by the path planning method.

avoidance experiment. From the result in Fig. 9(a) and (b), it was 4.3. Case 3: A person’s sudden appearance
confirmed that the robot with the proposed method can reach
its goal without colliding with the obstacles. Fig. 9(c) shows the We carried out the experiments in the illustrated environment
time history of the module activations. From the results, it was to verify the performance of the collision avoidance technique
confirmed that the robot approaches the obstacles, selects the T1 when a person shows up from the blind spot suddenly. We
module, and starts the collision avoidance. Moreover, when the confirmed in the simulation study that the robot collides with
robot approaches an obstacle, the robot can reach the goal without human in this situation when using only the T1 module. From the
colliding with the obstacle by selecting the T2 module. result shown in Fig. 11(3), we confirmed that the T1 and T2 modules
are selected simultaneously when the robot recognizes the person.
4.2. Case 2: dynamic obstacles Thereby, the robot can reach the goal without colliding with the
suddenly appearing obstacle.
We carried out experiments in the illustrated environment to
verify the effectiveness of the proposed method in the case of
moving obstacles. The experimental results are shown in Fig. 10 4.4. Case 4: Experiments in the hospital
as well as Fig. 9. From the results in Fig. 10, it was confirmed that
the robot recognizes the moving human, selects the T1 module We carried out several experiments to verify the performance
and starts the obstacle avoidance 4 seconds after the start of the of the path tracking and collision avoidance in the hospital. Fig. 12
experiment. From the results in Fig. 10(c), we confirmed that the shows the map based on the LRF data which are collected when
T2 module is selected at once at about 4.3 s after the T1 module is manually joy-sticking the robot in the first floor of the hospital,
selected. However, we think that the selection was wrong because and the path produced by the path planning method. In this map,
of a sensor error. Afterwards, by selecting the T2 and T1 modules a black area shows the area in which there is no obstacle and
at the same time when approaching the moving human, the robot a white line shows the path produced by path planning. In the
can reach the goal without colliding with the human. experiment, point 1 is the start point and point 2 is the goal point.
898 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899

New Energy and Industrial Technology Development Organization

(NEDO) in Japan.

References

[1] S. Thrun, M. Bennewitz, W. Burgard, A.B. Cremers, F. Dellaert, D. Fox, D. Hahnel,

C. Rosenberg, N. Roy, J. Schulte, D. Schulz, MINERVA: a second-generation
museum tour-guide robot, in: IEEE International Conference on Robotics and
Automation, vol. 3, 1999, pp. 1999–2005.
[2] H. Asoh, Y. Motomura, F. Asano, I. Hara, S. Hayamizu, K. Itou, T. Kurita, T. Matsui,
N. Vlassis, R. Bunschoten, B. Kroese, Jijo-2: an office robot that communicates
and learns, IEEE Intelligent Systems 16 (5) (2001) 46–55.
[3] R. Philippsen, R. Siegwart, Smooth and efficient obstacle avoidance for a tour
guide robot, in: IEEE International Conference on Robotics and Automation,
vol. 1, 2003, pp. 446–451.
[4] B. Graf, M. Hans, R.D. Schraft, Mobile robot assistant, IEEE Robotics and
Automation 11 (2) (2004) 67–77.
[5] R. Bischoff, V. Graefe, HERMES — a versatile personal robotic assistant,
Proceedings of the IEEE 92 (11) (2004) 1759–1779.
[6] K.K. Gupta, S. Zucker, Toward efficient planning: the path-velocity decompo-
sition, International Journal of Robotics Research 5 (3) (1986) 72–89.
[7] S. Ratering, M. Gini, Robot navigation in a known environment with unknown
moving obstacles, Autonomous Robots 1 (2) (1995) 149–165.
[8] R.C. Arkin, Motor schema-based robot navigation, International Journal of
Robotics Research 8 (4) (1989) 92–112.
[9] J. Borenstein, Y. Koren, The vector field histogram – fast obstacle avoidance
for mobile robots, IEEE Transactions on Robotics and Automation 7 (3) (1991)
278–288.
[10] V.J. Lumelsky, E. Cheung, Real time obstacle collision avoidance in teleoperated
whole sensitive robot arm manipulators, IEEE Transactions on Systems, Man,
and Cybernetics 23 (1) (1993) 194–203.
[11] D. Fox, W. Burgard, S. Thrun, The dynamic window approach to collision
avoidance, IEEE Transactions on Robotics and Automation 4 (1) (1997) 1–23.
[12] G.N. DeSouza, A.C. Kak, Vision for mobile robot navigation: a survey, IEEE
Transactions on Pattern Analysis and Machine Intelligence 24 (2) (2002)
237–267.
[13] O. Khatib, Real-time obstacle avoidance for manipulators and mobile robots,
International Journal of Robotics Research 5 (1) (1986) 90–98.
[14] Rodney A. Brooks, A robust layered control system for a mobile robot, IEEE
Journal of Robotics and Automation 2 (1) (1986) 14–23.
Fig. 13. Trajectories of the robot in the hospital experiments.
[15] S.S. Ge, Y.J. Cui, Dynamic motion planning for mobile robots using potential
field method, Autonomous Robots 13 (3) (2002) 207–222.

The length between the start and goal points is 12 m. The velocity
of the robot is 0.5 m/s. Fig. 13 shows the trajectory of the robot. In
Masaki Takahashi (Member, IEEE) received B.E. and M.E.
the experiments, some people obstruct the progress of the robot degrees in the Department of System Design Engineering
at random in the space in front of an elevator and in the narrow from Keio University in 2000 and 2002, respectively.
aisle. From the experimental results in Fig. 13(a)–(c), we confirmed He received the D. Eng. degree from Keio University,
Yokohama, Japan, in 2004. In 2004, he started working
that the robot recognizes the human, does the collision avoidance, as a Research Assistant of the 21st Century COE Program:
recovers the path tracking, and can reach the goal point (point 2) ‘‘System Design: Paradigm shift from Intelligence to Life’’.
in all three cases. From 2005 to 2008, he worked as a Research Assistant
of the Department of System Design Engineering, Keio
University, Yokohama, Japan and became an Associate
5. Conclusions Professor in 2009. His primary research interests include
human-robot interaction, motion and vibration control, and sensor fusion. He is a
This study proposed an obstacle collision avoidance technique member of American Institute of Aeronautics and Astronautics, The Japan Society
of Mechanical Engineers and The Robotics Society of Japan.
for a wagon truck pulling robot which uses an omni-directional
wheel system as a safe movement technology. Moreover, this
paper proposed a method to reach the goal along a global path Takafumi Suzuki received his B.E. degree in the depart-
produced by path planning without colliding with static and ment of system design engineering from Keio University
dynamic obstacles. The method is based on virtual potential fields. in 2008. He is currently a second year master’s degree stu-
dent at Keio University. His research interests include mo-
In the method, several modules with different prediction times tion control and obstacle collision avoidance for an au-
are processed in parallel to change the robot response according tonomous transfer mobile robot. He is a member of The
to its relative velocity and position to the obstacle. The virtual Robotics Society of Japan.
force calculated from each potential field is used to generate the
velocity command for the robot. Some experiments were carried
out to verify the performance of the proposed method. The robot
can deal with emergency situations adequately by acting on several
modules simultaneously according to the situation. Moreover,
Hideo Shitamoto received his B.E. and M.E. degrees in
from the experimental results in the hospital it was confirmed that management engineering from Kansai University in 1996
the robot can move along its planned path, and reach the goal and 1998, respectively. He has worked at Muratec R&D
without colliding with static and moving obstacles. Center, Murata Machinery, Ltd., Kyoto, Japan. His research
interests include motion control and obstacle collision
avoidance for an autonomous transfer mobile robot.
Acknowledgements

This work was partly supported by the project ‘‘Development

Project for a Common Basis of Next-Generation Robots’’ from the
 M. Takahashi et al. / Robotics and Autonomous Systems 58 (2010) 889–899 899

Toshiki Moriguchi received his B.E. and M.E. degrees in Kazuo Yoshida (Member, IEEE) received B.E. and M.E. de-
mechanical engineering from Kyoto University in 1992 grees in the Department of System Design Engineering
and 1994, respectively. He has worked at Muratec R&D from Keio University in 1973 and 1975, respectively. He re-
Center, Murata Machinery, Ltd., Kyoto, Japan. His research ceived the D. Eng. degree from Keio University, Yokohama,
interests include mechanical and control design, and Japan, in 1978. In 1981, he started working as a Research
system integration of an autonomous transfer mobile Assistant of the Department of Mechanical Engineering,
robot. He is a member of The Robotics Society of Japan. Keio University, Yokohama, Japan. He became an Associate
Professor in 1984 and a Professor in 1994. From 1996 to
2008, He was a Professor of the Department of System De-
sign Engineering, Keio University and from 2001 to 2008,
he was a Vice-President of Keio University. From 2003 to
2007, he was a Leader of the 21st Century COE Program: ‘‘System Design: Paradigm
shift from Intelligence to Life’’. His primary research interests include control en-
gineering, motion and vibration control, design and dynamics, robotics. He died in
2008.