Wireless Group Manipulation of Autonomously Guided Mobile Robots for Smart Living Space Applications

M.-H. Chen, D. Gu, Y.-D. Fu, C.-H. Pi, K.-S. Ou, and K.-S. Chen Department of Mechanical Engineering, National Cheng-Kung University, Tainan, Taiwan

(Tel: +886-6-2757575 ext.62192; E-mail: kschen@mail.ncku.edu.tw )

Abstract: In this work, by integrating omni-wheel mobile robots with X-Bee communication protocol, Arduino control,

IR range finders, and CMOS camera, as well as wiimote multi-zone localization, tasks such as obstacle and collision

avoidances, following, autonomously movement, and indoor localization of group robots are implemented as the first

step toward an autonomously control of group robots for smart living space applications. In conjunction with hardware

design, novel algorithms are developed for successfully realizing and demonstrating these tasks. With these key issues

being solved, more realistic scenario can be designed for achieving the real group robot applications for indoor service

in the future.

Keywords: Autonomously Guided Motion Control, Mobile Robots, Arduino, X-Bee, Group Manipulation

1. INTRODUCTION To understand and utilize the advantage of group

movement and cooperation of biological system are long

term pursuit by researchers in different aspects. For

robotics and artificial intelligence fields, by coordinating

and cooperating of large individual robots, complicate

tasks can be fulfilled. In addition, through proper

networking and communication, these robots can

exchange their information and learned from peers and

the environment for tackling more complicated tasks,

which cannot be done by single or ungrouped robots.

Consequently, it is possible to gain particular advantages

by applying the bio-inspired nature into the field. As a

result, several research laboratories have been deeply

engaged into the study for autonomously manipulation of

robot groups for specific applications by investigating in

virtually all aspects of mobile robots including remote

motion control, navigation, and speech recognition.

Previously, Rooker and Birk [1] utilized wireless

communications between two robots to establish

environmental map for robot manipulations. The ability

to manipulate a robot group enables new applications for

automatic indoor landscape arrangement and human

machine interactions, or for the smart architectural

technology; it might require coordination and group

manipulation of multiple mobile robots. This work is

inspired by the behavior of natural biological animal

groups, in which the individual member could either have

their own intelligent and information to perform motion

and decision making based on their own information or to

exchange information and perform group motion for

finishing a cooperative task. The bio-mimic approach

could potentially leads to a large scale integration of

robotic members to form a massive group for performing

complicated tasks.

In order to pursue the goal mentioned above, several

important fundamental tasks must be established and

realized. These fundamental issues including: collision

avoidance, global localization, obstacles avoidance, and

group manipulations. For obstacle avoidance, most

widely used technique is to utilize the information

observed by external-mounted CCD cameras for

subsequent trajectory planning [2,3] or by sensor

mounted on robots themselves [4]. Both approaches have

their own advantages and disadvantages. For example, the

former approach can obtain the global status of the robot

group but it cannot mimic the autonomous behavior of

individual elements. On the other hand, the latter one can

SICE Annual Conference 2011September 13-18, 2011, Waseda University, Tokyo, Japan

PR0001/11/0000-2143 400 2011 SICE- 2143 -

provide information of individual robot for decision but

the lack of global coordination makes it difficult to

perform tasks related to global positioning efficiently.

As a result, in this work, we combine both approaches by

utilizing multi-zone wiimote localization technique

developed by us [5] for global information assessments

and by IR range finders and CMOS cameras for recognize

the nearby objects. Based on the hardware and the

subsequent software implementation, it is possible to

achieve basic ability such as recognition, following [6],

obstacle avoidance, and global positioning. With these

basic abilities, it is possible to demonstrate more

dedicated group autonomous manipulations similar to the

previous work in robotic fish [7] in smart live

applications.

2. SYSTEM SETUP Four omni-wheel type mobile robots designed and

built by us are shown in Figure 1. These omni-wheels are

driven by three S03T_STD servo motors (powered by a

battery set) with an(UNO MEGA2560) Ardunio control

card with a clock 16 MHz and a (XBee ZNet2.5) X-Bee communication protocol (based on IEEE 802.15.4 with

maximum data rate 250kpbs) for the basic motion and

communication ability. The Arduino card sends PWM

signal between 0 and 255 to control the wheel rotating

speed (i.e., and the global translating ( x , y ) and self-rotating velocities ( ) of those robots can be obtained by the following kinematics equation:

0

,

where R and L are the radii of the wheels and the robot,

respectively. 1 is the motion direction, 2 is 1 to indicate forward or revise rotations. g1 and g2 are velocity gains.

Furthermore, depends on the application, associated

control laws to govern the relationship between the

feedback information and the controlled are also

established for performing subsequent tasks.

(a) (b)

(c) (d)

Fig. 1 The omniwheel mobile robots (a) key units, (b) the

fleet, (c) schematic plot to show key parameters, and (d)

IR range finder and CMOS camera assembly

In addition, a mesh topology between the host and

these robots is established by utilizing the X-Bee protocol.

The host sends sets of string signals in broadcasting

manner, which are received and decoded by each robot to

recognize the specific parts for their own. The mesh

topology enables us to overcome the possible signal

attenuation and delay due to the presence of obstacles.

Next, based on different task planning, each robot

equips three sets (one main and two auxiliary sets) of

(Sharp GP2Y0A21) IR range finder shown in Figure 1d

for detecting nearby objects for obstacle avoidance. The

effective ranges for the main and these two auxiliary IR

range finders are 0.15-5m (for a scanning angle 170) and 10 80cm (for a scanning range 240), for coarse and fine range determination, respectively. These range

finders are scanning back and forth for detecting possible

obstacles in a wide range. A novel algorithm is also

developed to coordinate the information from IR range

finders and the scanning angles for determining optimal

trajectory for obstacle avoidance.

Furthermore, each mobile robot also equips different

design of IR LED patterns and a CMOS IR camera

detached from Nintendo wiimotes for sensing the distance

and orientation between omni-wheels and for the purpose

- 2144 -

of object recognition. Next, wiimotes and their

corresponding IR LEDS are also installed on the ceiling

and the robots for globally monitoring the absolutely

location of these robots. In together with the multi-zone

localization technique developed by us [5], this provides

the global position information for external supervisor on

future task planning and control. Finally, the interaction

between the robot group and the external human

supervisor is established between the host computer and

either the master robot (group mode) or all robots

(individual mode) by using a graphic user interface

written under a LabVIEW environment. The entire

functional block diagram is shown in Figure 2.

Fig. 2 The functional block diagram of the entire work.

3. EXPERIMENTALSETUPAND ALGORITHMS DEVELOPMENT

Various fundamental benchmark problems have been

experimentally demonstrated to allow us to evaluate the

overall performance and to examine the possible faults in

recognition and communications between objects, as well

as to refine the manipulation schemes. By such efforts,

the above mentioned tasks are successfully demonstrated.

The performance measured and lesson learned could be

very valuable for future large scale integration.

Obstacle avoidance:

As mentioned earlier, an algorithm is proposed to

perform the obstacle avoidance. The algorithm is briefly

illustrated in Figure 3, the mobile robot is initially motion

in a straight manner and the IR range finders scan with a

rotating speed approximately 33.33 rpm to continuously

search for possible obstacles. The scanned area was

divided into four zones and the coordinates of the

possible obstacles are (di, i) (i=1~4). If there are any di less than its pre-defined threshold value dti, the obstacle

avoidance mode is then activated. A weighting index pi

(i=1~4) is then assigned based on di and it reflects the

obstacle presence distributions. Based on pi and i, the algorithm determines the velocity gains g1and g2 and the

possible orientation modification for the next step. By

such a dynamic scanning manner, the trajectory of the

robots will be evaluated and updated after each motion

step. Previous work[4] used ultrasonic sensor for obstacle

detection but it cannot precisely determine the obstacle

location and the robots would be difficult for making

following up decisions. On the other hand, this approach

can precisely determine the obstacle location and

promptly response it during the next step.

Fig. 3 Brief flowchart of the obstacle avoidance

algorithm.

Following:

As mentioned earlier, each robot equips with a set of

IR LEDs and a CMOS IR camera with a resolution of

1024768 for detecting the presence of nearby robots. By recognizing the image pattern of IR LEDs, each robot can

distinguish its neighborhood. Also, by evaluating the

coordinates of these LEDs, it is possible to calculate the

relative distance and orientation between two robots and

performing following motions. A following algorithm is

developed and briefly illustrated in Figure 4. The first

phase is to search the master robot. We divide the visible

area of the CMOS camera to five zones, denoted as qi,

i=1~5. Meanwhile, the motor scanning angles are also

partitioned into three different sets j, j=1~3. Hence the

total visible area is organized as 15 zones and a weighting

index Cij is assigned for each zone. Once an object is

- 2145 -

detected in a zone, the corresponding index is switched

from 0 to 1. Depends on the distribution of Cij, three

possible motions can be performed, i.e., straight motion,

locally self-rotation, and curvilinear rotation. Robots are

therefore able to find their master and to adjust their

speed and orientation for following. Once the master has

been identified, the system automatically changes to the

following mode by reducing the scanning range for

tracking the master and for increasing the system

bandwidth.

Fig. 4 Brief flowchart of the following algorithm.

External control and group autonomous movement:

A well-established communication network is the key

for successful manipulation of robot groups. This can be

further divided into two categories. The first category

concerns the interaction between the host and individual

robots and the established X-Bee protocol allows the host

computer send signal in a broadcasting manner to all

robots. These robots then decode the data to extract the

motion commands for their own and subsequently to

drive the servomotors to achieve the goal. By this

approach and careful motion planning, an external hosted

group motion can be achieved. On the other hand, by

mimicking the animal group motion behavior, the second

category deals with group autonomously movement. In

this task, one robot is assigned as the master and the rest

are slaves. The master received the command from the

host computer or by its own decision to plan the motion

based on sensed information. On the other hand, those

slave robots then move with the master robot to form

specific motion patterns. By this approach, an

autonomously guided robot group motion can be realized.

Wiimote indoor localization:

In order to monitor the location of each mobile robot in

real time for evaluating the effectiveness of our

algorithms, the multi-zone wiimote localization scheme

previously developed by us [5] is adapted here. As shown

in Figure 5, two wiimotes are mounted on the ceiling (2.3

m above) to cover an effective area of 1.82.7 m2 with a spatial resolution of 1.75mm for monitoring the motion.

In addition to this original purpose, it is immediately

realized that by properly utilizing the obtained global

position information, it is possible to perform more

sophisticated applications such as precision positioning of

an entire robot group and interactions between two

independent groups of autonomously robot in the future.

Fig. 5 The setup of the wiimote localization system

4. EXPERIMENTAL RESULTS Obstacle avoidance:

As shown in Figure 6a, a maze-like passage is

designed to test the performance of the obstacle

avoidance. A robot is commanded to walk through the

passage within 10 seconds without collisions with the

walls. On the other hand, a path with a dead end is also

used and the results shown in Figure 6b indicate that the

robot can successfully walk out from it without the

concern of trapping inside. These features are important

for indoor service robot since passages could be

complicated and with many dead ends exist inside a

typical living space. Robots without the above mentioned

abilities would be impractical for indoor service

applications.

- 2146 -

(a)

(b)

Fig. 6 Obstacle avoidance demonstrations

Collision avoidance between two robots:

In a multi-robot workspace, collisions between two or

more robots must be prohibited and this issue is

investigated here. This can be treated as a dynamic

situation of the above obstacle avoidance problem. As

shown in Figure 7a, two mobile robots approach each

other. Once the relative distance is within the threshold,

both robots will make decisions to avoid collisions based

on the obstacle avoidance algorithm addressed above.

Following:

Following between a group of slave robots and a

master robot is a fundamental technical ability for

subsequent realization in autonomous motion control of

group robots. Figure 7b demonstrates the results of our

algorithm in robot following. The master robot performs a

general planar motion and followed by a slave robot. As

one can see, the slave robot follows the path of the master

robot very well.

(a)

(b)

Fig. 7 Interactions between two robots (a) collision

avoidance and (b) following motion

External group control:

In this experiment, the host broadcasts a massage to all

three mobile robots and asks them moves toward their

corresponding destinations. The results are successfully

shown in Figure 8 and the result indicates that it is

possible to simultaneously control a large number of

robots for real applications (such as smart building

elements) by just sending an information package while

the bandwidth can still be maintained. This would be very

useful for smart building applications where a large

number of building element must be moved to specific

locations for fulfilling the functional requirements.

Fig. 8 An external group control demonstration

Autonomously movement:

Finally, autonomously movement experiments are

performed based on the above established technical

capabilities. Four our mobile robots are designed to

perform three tasks. First, we allow these four robots

moves freely with collision free in the workspace as

shown in Figure 9a. Next, as shown in Figure 9b, with

one robot serves as the master, all other slave robots

moves toward the master once they are asked. Finally,

these slave robots follow the master to form a following

- 2147 -

motion without collisions as shown in Figure 9c. The

successful demonstrations shown allow us to investigate

more realistic situation for realizing the real group robot

applications for indoor service in the future.

(a)

(b)

(c)

Fig. 9 Autonomously movement demonstration

5. SUMMARYAND CONCLUSION Autonomous control of a group of mobile robots has

the fundamental importance in bio-mimic or smart living

space related applications. However, several fundamental

concerns in communication, coordination, obstacle

avoidance, and cooperation must be solved and

demonstrated before detail task planning. In this work,

both hardware design and software algorithm

implementation are developed for establishing the

associated fundamental capability. In particular, this work

proposes a novel obstacle avoidance architecture and its

associated search algorithm to dynamically search

possible obstacles and update the moving trajectory in

real time. In addition, by integrating this scheme with the

global position information provided by the wiimote

multi-zone localization, it is expected more sophisticated

schemes such as the interaction between two

autonomously controlled robot groups can be further

developed in the future for indoor service applications..

ACKNOWLEDGEMENT This work is supported by National Science Council

under contact numbers: NSC97-2221-E-006-152-MY3

and NSC 99-2221-E-006-173-MY2.

REFERENCES

[1] M. N. Rooker and A. Birk, "Multi-robot exploration

under the constraints of wireless networking,"

Control Engineering Practice, Vol.15, pp.435-445,

2007

[2] C. L. Hwang and L. J. Chang, "Trajectory tracking

and obstacle avoidance of car-like mobile robots in

an intelligent space using mixed H2/H

decentralized control," IEEE/ASME Trans. on

Mechatronics, Vol. 12, pp.345-352, 2007

[3] A. H. Pratomo et al., "Position and Obstacle

Avoidance Algorithm in Robot Soccer," Journal of

Computer Science, Vol. 6, pp. 173-179, 2010.

[4] Y. S. Chen and J. G. Juang, "Intelligent Obstacle

Avoidance Control Strategy for Wheeled Mobile

Robot," ICROS-SICE International Joint

Conference, Fukuoka, Japan, August 18-21, 2009

[5] P. W. Chen et al., "IR indoor localization and

wireless transmission for motion control in smart

building applications based on Wiimote

technology," in Poceedings of SICE Annual

Conference 2010, pp.1781-1785, Taipei, Aug 2010.

[6] B. S. Choi et al., "Target-following scheme for

mobile robots based on the virtual sub-goal points,"

Information, Vol.13, pp. 1929-1938, 2010.

[7] Y. Hu , et al. , "Underwater target following with a

vision-based autonomous robotic fish," American

Control Conference2009,pp.5265-5270, St. Louis,

MO,USA, 2009.

- 2148 -