Sunteți pe pagina 1din 4

Autonomous Robotic Sensor Agents

Emil M. Petriu ( I ) , Thom E. malen('), and Rami Abielmona

University of Ottawa, Ottawa, ON Canada


(I)
(petriu, rabielmo)@ite.tiottawa.ca
(2) Communications Research Centre Canada, Ottawa, ON Canada
thom.whale@t crc.ca

Abstract Integrating the partial and heterogeneous sensor-views of


the environment into a coherent model that is easy-to-use for
This paper discusses development aspects of a network of computer controllers and human operators alike is another
intelligent wireless autonomous Robotic Sensor Agents de-
plcyed in the field for active investigation of complex envi- important aspect to be considered.
ronmental parameters.
Keywords - intelligent sensors, mobile agents, distributed
sensor networks, wireless sensors, multi-sensor data jiuion,
behavior-based robots.

1. Introduction

This paper revisits the development principles recently


proposed by authors in [l] and [2] and reports new imple-
mentation aspects for a network of autonomous wireless
Robotic Sensor Agents ( R S A ) for complex environment
monitoring currently under development at the Sensing and
Modeling Research Laboratory, in SITE, at the University of
Ottawa. These RSAs are intended for autonomous investi-
gation of relevant parameters in natural living environments,
industrial or laboratory hazardous environments, polluted . .I
I
w ! I

environments, water treatment plants, nuclear stations, war


zones, or remote environments such as mines. WIRELESS COMMUNICATIONNETWORK
Monitoring is done by continuously collecting sensory
data from stationary and mobile RSA deployed in the field,
as illustrated in Figure 1. These data are then hsed in a vir-
tualized reality model of the explored environment, which
VIRTUALIZED REALITY
can be used directly or made remotely available to human MODEL OF THE EXPLORED
supervisors for model-based interactive control applications.
The type of sensors, their number and bandwidth, the dif-
ficulties encountered while deploying the sensors in the
field, the reliability of the sensory data and of the data c o n f-igure 1. Distributed wireless network of mobile and
munication network, are all requirements that have to be stationary intelligent robotic sensor agents
considered when designing the robotic sensor agents and the
deployed in the environment
distributed sensor network architecture.

0-7803-8109-2/ 03 / $17.000 2003 IEEE


2. Autonomous Robotic Sensor Agents RSAs should be capable of selective environment percep-
tion focusing on parameters that are important for the spe-
Autonomous RSAs are not functionally and operationally cific task and avoid wasting resources on processing irrele-
identical. A variety of experimental autonomous robot sen- vant data. Different sensor planning strategies are used for
sor agents, illustrated in Figure 2 and Figure 3, are currently the placement of the fixed and mobile sensor agents in such
developed to cover all four perception phases of the envi- a way as to get optimum performance during specific sens-
ronment parameters: far away, near to, touching, and mi- ing tasks and for the real-time selection of sensing opera-
nipulation. tions to minimize the observed system entropy, [3] and [4].
A multisensor data fusion mechanism is used for the in-
tegration of heterogeneous sensory data into composite
models of 3D object shape, surface and material properties,
heat transfer and radiation (EM, thermal, radioactive, opti-
cal, etc.) characteristics. The multi-sensor fusion framework,
[ 5 ] , deals in a consistent way with a diversity of measure-
ment data produced by RSAs. Such a multi-sensor fusion
system has to:
(i) organize data collection and signal processing from
different types of sensor,
(ii) produce local and global world models using the
multi-sensor information about the environment,
(iii) integrate the information from the different sensors
into a continuously updated model of the system.
In many applications, there is a considerable communica-
tion delay and randomness between the information col-
lected from the field deployed sensor agents. This may d-
fect the fidelity and consistency of the integral world model
of the monitored environment, [6] and [7].
Figure 2. Two-wheel robotic platform for an We are using the “time clutch” and “position clutch”
experimental mobile RSA equipped concepts proposed in [6] to maintain the coherence &tween
with wireless camera and IR sensors the information acquired from different sensors agents
measuring parameters of the same object cr environment
region. In order to avoid fatal errors and reduce the effect of
the communication &lay, we are using a distributed virtual
environment allowing for the maintenance of a shared world
model of the physical environment that is explored, [8].

2.1. Intelligent autonomous agent behavior

In order to provide the high degree of autonomy required


by their functions, RSAs have onboard sensing, computing,
and wireless communication capabilities as well as the abil-
ity to autonomously perform exploratory motions, as illus-
trated in Figure 4.
Leaming allows autonomous robots to acquire knowl-
edge by interacting with the environment and subsequently
adapting their behavior. Behavior learning methods are used
to solve complex control problems that autonomous robots
would encounter in an unfamiliar real-world environment.
Neural networks, fuzzy logic, reinforcement- and evolution-
ary-leaming methods can be used to implement basic behav-
Figure 3. Two-wheel robotic platform for an ioral functions, [9], [lo], and [ 111.
experimental mobile RSA equipped We are investigating the use of Brooks’ reactive-
with sonar and 2-DOF robot arm behavior paradigm, [9], as an alternative to the traditional
with tactile sensors in its fingers. function oriented sensing-planning-acting control strategies,
[12]. The behavior paradigm is based on a task-wise decom-
position of the control functions in special-purpose simple
task-achieving modules. We are verifylng claims that neither for the upgrading of the operational capability of surviving
strategic planning nor carefdly calibrated sensors are neces- agents.
sary to produce robust intelligent reactive -behavior in
autonomous robotic agents, [9] and [ 101.

_ _ _ _ _ _ _ _ _ _ _ _ ,' COMMUNICATION
I
MULTCCARRIER '

HIGHLEVEL '\ MECHANISMS ,


INFORMA TlON
EXCHANGE
*'-
-
;'-- - -*.:-
MECHANISMS: I'
' o'STRIBUTEo
XML. Grammars.... s'
;
.* --------__- FRAMEWORK.
COMPUTING
,
fs

OEL~BERATIVEIREACTIVE
~ CONTROL
---- _ _ _ _--_ -
-,,) )
Figure 4. Onboard control system for an Figure 5. Hexagonal-shaped modular RSA
autonomous wireless robotic platform able to couple with other
sensing agent similar RSAs.

2L!. Social behavior of RSA An intriguing and controversial question, [19], is the
value of extending the recycle-the-dead (.i.e. the useless)
All RSAs are by definition instinctive information seeking cannibalism of RSAs to a more aggressive
agents. When the costs of deploying sensors agents is pro- Iiigfish-eats-smallerfish survival of the fittest behavior.
hibitive, they would benefit from having survival behav- Obviously, such an aggressive survival behavior should not
iorhstinct, cooperation skills, adaptation and learning allow for suicidal actions. An agent should not prey on
abilities, [ 131, [ 141, and [151. Like a human being, each RSA equally strong or more powerful RSAs. While such behav-
as an intelligent "bot ", [ 161, may have it's own personality, ior may evolutionary lead to more individually efficient
[l'l]. ISAS, it may affect the global mission by decimating the
Cooperating agents should be able to work together with population of agents deployed in the environment. It is quite
other RSAs toward the overall goal, which is to maximize possible that fewer more operationally efficient, from an
the: information acquired from the environment. This coop- information-gathering standpoint, RSAs will not do a better
eraiion should also allow modular RSAs to permanently or global job than more albeit not so efficient, but nevertheless
temporarily couple forming new structures better adapted to .still operational agent.
solving specific problems. We are currently evaluating the In order to facilitate recycling, each RSA has a status ad-
performance of an experimental RSA platform with a hex- vertising mechanism telling other agents about its job related
agonal shape shown in Figure 5. junctional qualification and health level.
In order to make possible the implementation of these
characteristics, R S A have modular reconfigurable structures 3. Networking Robotic Sensor Agents
with accessible and easy to assembly/disassembly compo-
nents. In order to provide a flexible extensible open mechanism
In order to allow for i-bot species survival, [181, RSAs for interoperability, an agent-based resource management
should be able to cannibalize/recycle other agents that are framework should address the functional and communica-
opixationally dead, which otherwise will be abandoned in tion needs of each RSA.
the field. Providing RSAs with such a behavior would allow
Heterogeneous RSAs cannot realistically be expected to E. Tunstel, “Behavior hierarchy for autonomous mobile IO-

talk exactly the same language. However, they will share bots: Fuzzy-behavior modulation and evolution,” Intl. Jour-
nal ojrntelligent Audomation and Soft Computing, Special Is-
domain-specific knowledge, which may be expressed by sue: Autonomous Control hgineering, Vol. 3, No.1, pp. 37-50,
each of them in a different dialect. Accordingly, the com 1997.
munication management framework should define a domain P. Rusu, E. M. Petriu, T.E. Whalen, A. Comell, and H. J.W.
Spoelder, “Behavior-Based Neurc-Fuzzy Controller for Mo-
specific semantic for common knowledge and functions. bile Robot Navigation,” Proc. IMTC/2002, IEEE Instrum.
This framework is expected to act as a universal translator Meas. Technol. ConJ, pp. 1617-1622, Anchorage, AK, USA,
between speakers of different dialects, [20]. May 2002.
R.J. Gallimore, N.R. Jennings, H.S. Lamba, C.L. Mason, B.J.
Orenstem, “Cooperating Agents for 3D Scientific Data Inter-
pretation,” IEEE Trans. Syst. Man Cybern. -Part C: Applica-
Acknowledgments tions andReviews, Vo1.29, No.1, pp.110-126, 1999
L. E. Navarro-Serment, R. Grabowski, C. J.J. Paredis & P.
This work was funded in part by the Natural Sciences K. Khosla, “Heterogeneous Teams of Modular Robots for
Mapping and Exploration”, Technical report ICES 04-1 1-99,
and Engineering Research Council of Canada (NSERC), the The Institute for Complex Engineered Systems, Camegie
Communications Research Centre Canada (CRC), and the Mellon University, Pittsburgh, PA, USA, 1999.
Communications and Information Technology Ontario A. Banerjee, T. DeFanti, A. Hudson, B. Dodds, and J.R.
(CITO). The authors gratefully acknowledge the assistance Curtis, “A Behavioral Layer Architecture for Telecollabora-
tive Virtual Manufacturing Operations,” IEEE Trans. Robot-
of Alan Stewart in the implementation of the robotic sensor ics Automat, Vol. 16, No. 3, pp. 218-227, 2000
platforms. A. Leonard, Bots - The Origin of New Species, Hardwired,
San Francisco, 1997.
S. Pinker, How The Mind Works, W.W. Norton & Co., New
York, 1997.
R. Dawkins, The Selfish Gene-New Edition, Oxford Univer-
References sity Press, Oxford, 1989.
M. Crichton, Prey, HarperCollins Publishers, New York,
E.M. Petriu, G.G. Patry, T.E. Whalen, A.H. AI-Dhaher and 2002.
V. 2 . Groza, “Intelligent Robotic Sensor Agents for Envi- E.M. Petriu, N.D. Georganas, D.C. Petriu, D. Makrakis, and
ronment Monitoring,” Proc. VIMS’O2,IEEE Intl. Symposium V.Z. Groza, ”Sensor-Based Information Appliances,” IEEE
on Virtual and Intelligent Measurement Systems, pp. 14- 19, Instrum. Meas. Mag., Vol. 3, No. 4, pp. 31-35, 2000.
Girdwood, AK, USA, May 2002
R. Abielmona, E.M. Petriu, V. Groza, T.E. Whalen, “Can an
Intelligent Society of Robots Survive in a Hostile Environ-
ment?,” Proc. 2003 Canadian Con$ Elec. Comput. Eng.,
Montreal, Que., May 2003.-
G. Roth, M.D. Levine, “Minimal Subset Random Sampling
for Pose Determination and Refinement,” in Advances in
Machine Vision, (C. Archibald, E. Petriu, Edit.), pp.1-21,
World Scientific, 1992.
K.A. Tarabanis, P.K.Allen, R.Y. Tsai, “A Survey of Sensor
Planning in Computer Vision,” IEEE Trans. Robotics Auto-
mat., Vol.ll,No.l,pp.86-104, 1995.
R. Joshi, A.C. Smderson, Multisensor Fusion: A Minimal
Representation Framework, World Scientific, 1999.
L. Conway L., R.A. Voltz., and M.W. Walker, “Teleautonc-
mous Systems: Projecting and Coordinating Intelligent Ac-
tion at a Distance,“ IEEE Trans. Robotics Automat., vol. 6,
no. 2, pp. 146-158, 1990.
G.V. Kondraske, R.A. Votz, D.H. Johnson, D. Tesar, J.C.
Trinkle, and C.R. Price, “Network-Based Infrastructure for
Distributed Remote Operations and Robotics Research,”
IEEE Trans. Robotics Automat, Vol. 9, No. 5, pp. 702-704,
1993.
F.C.A. Groen, N. Vlassis, M. Boasson, H.J.W. Spoelder, and
H.E. Bal, “A framework for maintaming a shared world
model in dynamic environments between differentiated an-
bedded systems and allowing interaction with human super-
visors,” in Proc. bd PROGRESS Workshop on Embedded
Systems, Veldhoven, The Netherlands, 2001.
R. Brooks, “A Robust Layered Control System for a Mobile
Robot,“ IEEE J. Robotics Automat., v01.2, no.1, pp. 14-23,
1986.
R.C. Arkin, Behavior-Based Robotics, The MIT Press, 1998.

S-ar putea să vă placă și