CONTRIBUTI SCIENTIFICI

A RTIFICIAL INTELLIGENCE AND ROBOTICS

ANTONIO CHELLA
Dipartimento di Ingegneria Informatica - Università degli Studi di Palerm o

LUCA IOCCHI
Dipartimento di Informatica e Sistemistica - Università degli Studi di Roma “La Sapienza”

IRENE MACALUSO
Dipartimento di Ingegneria Informatica - Università degli Studi di Palerm o

DANIELE NARDI
Dipartimento di Informatica e Sistemistica - Università degli Studi di Roma “La Sapienza”

1 Introduction to build intelligent robots are clearly identified by the re-

search community, and the developement of robots is again
Artificial Intelligence and Robotics have a common root viewed as a prototypical case of AI system [29]. Following
and a (relatively) long history of interaction and scien- the title of the paper we shall refer to this body of research
tific discussion. The birth of Artificial Intelligence and as AI Robotics.
Robotics takes place in the same period (’50), and initially We conclude this brief introduction with a disclaimer:
there was no clear distinction between the two disciplines. the views presented in the paper are those of AI research,
The reason is that the notion of “intelligent machine” nat- that use robots as a preferred model of intelligent agent
urally leads to robots and Robotics. One might argue that and there is no attempt to provide a comprehensive survey.
not every machine is a robot, and certainly Artificial In- In the recent years, Robotics researchers have also tackled
telligence is concerned also with virtual agents (i.e. agents some of the issues that are dealt with in the present paper,
that are not embodied in a physical machine). On the other but the view of Robotics research towards Artificial Intel-
hand, many of the technical problems and solutions that ligence may not be properly reflected in the paper.
are needed in order to design robots are not dealt with by The paper is organized as follows. In the next section
Artificial Intelligence research. we address the major scientific issues in the field. Then we
A clear separation between the fields can be seen in look at the connections and relationships with other topics
the ’70, when Robotics becomes more focused on indus- addressed in this collection, and with other disciplines. Af-
trial automation, while Artificial Intelligence uses robots terwords, we present some application scenarios that have
to demonstrate that machines can act also in everyday en- been developed by the research in Italy.
vironments.
Later, the difficulties encountered in the design of
robotic systems capable to act in unconstrained environ- 2 Research issues
ments led AI researchers to dismiss Robotics as a preferred
In this section we analyse the recent work which can be
testbed for Artificial Intelligence. Conversely, the research
characterised as AI Robotics, by arranging it into the two
in Robotics led to the development of more and more so-
basic issues in robot design: Action and Perception.
phisticated industrial robots.
This state of affairs changed in the ’90s, when robots be-
gun to populate again AI laboratories and Robotics specif- 2.1 Action
ically addressed also less controlled environments. In par- While there is nowadays a general agreement on the basic
ticular, robot competitions1 started: indeed they played a structure of the autonomous agent/robot, the question of
major role in rehestablishing a strict relationship between how this structure can be implemented has been subject to
AI and Robotics, that is nowadays one of the most promis- a long debate and is still under investigation.
ing developments of research both in the national context
Agents and, specifically, robots, usually present various
and at the European level.
kinds of sensing and acting devices. The flow of data from
Summarizing, the borderline between the work in Arti- the sensors to the actuators is processed by several different
ficial Intelligent and Robotics is certainly very difficult to modules and the description of the interaction among these
establish; however, the problems to be addressed in order modules defines the agent s architecture.
1 See, for example, AAAI robot competitions and challenges The first, purely deliberative, architectures [12, 22] view
(www.aaai.org) and RoboCup competitions (www.robocup.org) the robot as an agent embedding a high-level representa-
Anno III, N° 1/2, Marzo-Giugno 2006 87
 CONTRIBUTI SCIENTIFICI

tion of the environment and of the actions that it can per- Action theories A number of theories of actions have
form. Perceptual data are interpreted for creating a model been developed in order to represent the agent’s knowl-
of the world, a planner generates the actions to be per- edge. They are characterized by the expressive power, that
formed, and the execution module takes care of executing is the ability of representing complex situations, by the
these plans. In practice a sense-plan-act cycle is repeatedly deductive services allowed, and by the implementation of
executed. The problem is that building a high-level world automatic reasoning procedures. Several formalisms have
model and generating a plan are time consuming activities been investigated starting from Reiter’s Situation Calculus
and thus these systems have shown to be inadequate for [27, 13]: A-Languages (e.g., [14]), Dynamic Logics (e.g.,
agents embedded in dynamic worlds. [11]), Fluent and Event Calculi (e.g., [8]).
Reactive architectures focus on the basic functionalities The proposed formalisms address several aspects of ac-
of the robot, such as navigation or sensor interpretation, tion representation including sensing, persistence, non-
and propose a direct connection between stimuli and re- determinism, concurrency. Moreover, they have been fur-
sponse. Brooks’s subsumption architecture [4] is com- ther extended with probabistic representations, represen-
posed by levels of competence containing a class of task- tations of time etc. However, much of the work carried
oriented behaviors. Each level is in charge of accomplish- out on action theories has been disconnected from appli-
ing a specific task (such as obstacle avoidance, wander- cations on real robots, with some notable exceptions. (see
ing, etc.) and the perceptual data are interpreted only for for example [5, 3, 7, 11]). A more popular approach to
that specific task. Reactive architectures, while suitably action representation on robots is based on decision mak-
addressing the dynamics of the environment, do not gener- ing techniques, which maximise the utility of the actions
ally allow the designer to consider general aspects of per- selected by the robot, depending on the operational con-
ception (not related to a specific behavior), and to identify text [29]. However, this approach does not provide an ex-
complex situations. In fact, the use of a symbolic high- plicit representation of the properties that characterize the
level language is not possible, since it would necessarily dynamic system, while focussing on the action selection
require building a world model, and thus reasoning is usu- mechanism.
ally compiled into the structures of the executing program.
The lack of previsions about the future limits these systems
in terms of efficiency and goal achievement. Architectures There are many features that are consid-
The above considerations led to a renewed effort to com- ered important in the design of agents’ architectures and
bine a logic-based view of the robot as an intelligent agent, each proposal describes a solution that provides for some
with its reactive functionalities. To this end a new research of these features. Approaches to architectures that try to
field is developing in the last years: Cognitive Robotics. combine symbolic and reactive reasoning are presented for
The name was first introduced by the research group at the example in [1, 26] as so called Hybrid Architectures. We
University of Toronto led by Ray Reiter [19]. The most can roughly describe a layered hybrid architecture of an
recent view of cognitive robots, that has been accepted, for agent with two levels: the deliberative level, in which a
example in the EU framework, certainly keeps the origi- high-level state of the agent is maintained and decisions on
nal goal of embedding a reasoning agent into a real robot, which actions are to be performed are taken, and the op-
but also takes a more general perspective, by looking at the erative level, in which conditions on the world are verified
perception/action cycle in a broader sense, in bio-inspired and actions are actually executed.
systems, as well as in the work on recognition and genera- The embodied intelligence approach generalizes
tion of emotional behaviours (see next section). Cognitive Brooks’s ideas (see e.g., [32], [25]). The robot is a real
Robotics aims at designing and realizing actual agents (in physical agent tightly interacting with the environment and
particular mobile robots) that are able to accomplish com- the robot behavior is generated not by the robot controller
plex tasks in real, and hence dynamic, unpredictable and alone, but it emerges by means of the interactions between
incompletely known environments, without human assis- the robot with its body and the environment.
tance. Cognitive robots can be controlled at a high level, Other contributions to the realization of robot architec-
by providing them with a description of the world and ex- tures come from evolutionary computing, where evolution-
pressing the tasks to be performed in the form of goals to ary robotics is a research field aiming at developing robots
be achieved. through evolutionary processes inspired by biological sys-
The characterizing feature of a cognitive robot is the tems [23]. For example, neuro-fuzzy systems have been
presence of cognitive capabilities for reasoning about the successfully used in the design of robot architectures.
information sensed from the environment and about the ac- Often, the work on architectures is developed in the con-
tions it can perform. The design and realization of cogni- text of robot programming environments, including ad-hoc
tive robots has been addressed from different perspectives, specialised control languages. Most of this work is more
that can be classified into two groups: action theories and concerned with engineering aspects and will not be ad-
system architectures. dressed here.

88 Anno III, N° 1/2, Marzo-Giugno 2006

 CONTRIBUTI SCIENTIFICI

2.2 Perception represents schematic and instantiated knowledge about dy-

namic scenes. This conceptual formalism mediates be-
Robot perception is a prominent research field in AI and tween the spatiotemporal geometric descriptions extracted
Robotics. Current robotic systems have been limited by by video cameras and the high-level system for the gener-
visual perception systems. In fact, robots have to use other ation of natural language text.
kinds of sensors such as laser range finder, sonar, and so A related system [6] is based on three level of repre-
on in order to bypass the difficulties of vision in dynamic sentations: the subconceptual, the conceptual and the sym-
and unstructured environments. bolic level. In particular, the main assumption is that an in-
A robotic agent acting in the real world has to deal with termediate representation level is missing between the two
rich and unstructured environments that are populated by classes of representations mentioned above. In order to fill
moving and interacting objects, by other agents (either this gap, the notion of conceptual space is adopted, a rep-
robots or people), and so on. To appropriately move and resentation where information is characterized in terms of
act, a robot must be able to understand the perceptions of a metric space. A conceptual space acts as an intermedi-
the environment. Understanding, from an AI perspective, ate representation between subconceptual (i.e., not yet con-
involves the generation of a high-level, declarative descrip- ceptually categorized) information, and symbolically orga-
tion of the perceived world. Developing such a description nized knowledge.
requires both bottom-up, data driven processes that asso- Some basic primitives (Find, Track, Reacquire) that de-
ciate symbolic knowledge representation structures with fine the anchoring of symbols in sensory data as a prob-
the data coming out of a vision system, and top-down pro- lem per se and independent of any specific implementation
cesses in which high-level, symbolic information is em- have been proposed and discussed [9].
ployed to drive and further refine the interpretation of the In order to define a more general logical account of robot
scene. perception linking sensory data to high-level representa-
To accomplish its tasks, a robot must be endowed with tion, recently an abductive theory of perception has been
selective reasoning capabilities, in order to interpret, clas- proposed [31]. In this theory, the task of robot perception
sify, track and anticipate the behavior of the surrounding is to find and explanation of sensory data according to a
objects and agents. Such capabilities require rich inner background theory describing the robot interactions with
representations of the environment firmly anchored to the the environment.
input signals coming from the sensors. In other words, the
meaning of the symbols of the robot reasoning system must
be anchored in sensorimotor mechanisms. 3 Interaction with other AI elds
On the one side, the robot vision community approached As already mentioned, the research on AI Robotics inter-
the problem of the representation of scenes mainly in terms sects a number of subfields of AI. Indeed, the robotic agent
of 2D/3D reconstruction of shapes and of recovery of their can be seen as a main target for the grand goal of Artificial
motion parameters, possibly in the presence of noise and Intelligence, and thus for all the aspects of AI somewhat re-
occlusions, in order to control the motion of the robot. This lated to Robotics. Below, we address the main connections
approach is known as visual servoing of robot system [10]. with the other AI research topics included in this collec-
On the other side, the AI community developed rich and tion.
expressive formalisms for image interpretation and for rep-
resentation of processes, events, actions and, in general, of Machine Learning Learning approaches are being ap-
dynamic situations, as mentioned in the previous section. plied to many problems arising in the design of
However, the research on robot vision and on AI knowl- robots. According to the structure adopted above,
edge representation evolved separately, and concentrated both action and perception can be supported by learn-
on different kinds of problems. On the one hand, the robot ing approaches. Moreover, several approaches that
vision researchers implicitly assumed that the problem of include a training step are pursued ranging from ma-
visual representation ends with the 2D/3D reconstruction chine learning approaches to genetic programming,
of moving scenes and of their motion parameters. On the and neural networks.
other hand, the AI community usually did not face the
problem of anchoring the representations on the data com- From the standpoint of action, learning approaches
ing from sensors. can be used for the basic action skills, specifically lo-
Starting from the seminal paper of Reiter and Mack- comotion, but also learning cooperative behaviours,
worth [28], some proposal have been made in this research adaptation to the environment, and learning oppo-
field, a few of them briefly described below. nents’ behavior, among others.
The main steps toward an effective cognitive vision sys- Obviously, the learning process must face the chal-
tem for dynamic scene interpretation have been recently lenges of the experiments with real robots. Neverthe-
discussed [20] by adopting a fuzzy metric temporal Horn less, in several experimental settings (e.g. RoboCup),
logic in order to provide an intermediate formalism that learning and adaptation of the basic skill, such as

Anno III, N° 1/2, Marzo-Giugno 2006 89

 CONTRIBUTI SCIENTIFICI

walking, vision calibration, have shown to be much are needed: physics, electrical engineering, electronic en-
more effective than parameter tuning by hand. gineering, mechanical engineering, computer science, AI,
and so on. It is therefore difficult also to have a common
Edutainment Toy robots are very promising to be used background of terms, notations and methodologies. In this
both for research purposes and for education, because sense, the efforts to define a common ontology of terms for
of low costs and high attraction for students. Even a robotics science [15] are noteworthy.
though, at this moment, the available educational kits In particular, AI Robotics interacts with several research
seem to provide too limited capabilities, toy robots disciplines outside AI.
are certainly an interesting commercial market. Con-
sequently, the design of intelligent toy robots is an Industrial Robotics Many contact points may be found
interesting opportunity for AI researchers. between AI, Robotics and Industrial Robotics. In
early days there were not clear and cut distinctions
The experience with Aibo robots [33] shows this po-
between the two fields, as already mentioned. Today,
tential: they have been successfully used by many re-
research in Industrial Robotics is oriented towards the
search groups in the world not only in the RoboCup
safe and intelligent control of industrial manipulators
competitions (Four-Legged League), but also for
and in the field of service robotics. The methodolo-
demonstrating other AI and Robotics research issues.
gies in Industrial Robotics are grounded in Automatic
Multi agent systems A multi-robot system (MRS) can be Control Theory [30]. The relationship between the
considered as a multi-agent system (MAS), but the robot and the environment is generally modeled by
techniques for achieving coordination and coopera- means of several types of feedback systems. More-
tion in MAS are often not well suited to deal with the over, methodologies are typically based on numerical
uncertainty and model incompleteness that are typi- methods and optimization theory.
cal of Robotics. Multiple robots may achieve more Computer Vision Robot Vision is specific with respect
robust and more effective behavior by accomplish- to computer vision, because Robot Vision is intrin-
ing coordinated tasks that are not possible for single sically active, in the sense that the robot may actively
robots. Groups of homogeneous and heterogeneous find its information sources and it can also reach the
robots have a great potential for application in com- best view position to maximize the visual informa-
plex domains that may require the intelligent use and tion. Moreover, Robot Vision must be performed in
merge of diverse capabilities. The design, implemen- real-time, because the robot must immediately react
tation, and evaluation of robots organized as teams to visual stimuli. In general, the robot cannot pro-
pose a variety of scientific and technical challenges. cess for a long time the same image because the en-
vironmental conditions may vary, so the robot has to
Natural Language Processing It is an obvious require-
deal with approximate, but just in time information.
ment of home and service robotics the ability to in-
Several research topics and debates in this field have
teract with people in natural language; therefore, nat-
strong correlations with AI and Robotics, for exam-
ural language processing techniques find an interest-
ple, if a Computer Vision system may be based on
ing application domain on robots (see for example the
inner representation of the environment or it should
RoboCare project below).
be purely reactive.
Logics for AI and Automated Reasoning The connec- Mechatronics Mechatronics encompasses competencies
tion to the Logics for AI and Automated Reasoning from electrical engineering, electronic engineering,
is central to the work on Cognitive Robotics, but we mechanical engineering. All of these competencies
do not further expand it here, as it is discussed in the are strictly related to AI and Robotics: the research
previous section. field of electrical engineering concerns motors and ac-
Evolutionary Computation and Genetic Programming tuators, while electronic engineering mainly concerns
Evolutionary Robotics is a new approach that looks boards for robot control, for data acquisition and in
at robots as autonomous artificial organisms that general for the hardware that makes the robot oper-
develop their own skills in close interaction with the ational. Mechanical engineering concerns of course
environment without human intervention. Evolution- the mechanical apparatus of the robot itself. From this
ary robotics thus applies techniques coming from point of view, Mechatronics, AI and Robotics have
evolutionary computation. tight relations: Mechatronics mainly focuses on the
robot hardware at all levels, while AI and Robotics
take care of the software that makes the robot opera-
4 Interaction with other disciplines tive and autonomous.
Robotics is a multidisciplinary field: to make an opera- Embedded Systems The AI software architecture of a
tional robot, several contributions from many disciplines robot is naturally embedded into the physical appara-

90 Anno III, N° 1/2, Marzo-Giugno 2006

 CONTRIBUTI SCIENTIFICI

tus of the robot. Therefore, the robot software system in particular global vision as well as other external sens-
needs to work in real time in order to guarantee that ing devices are not available. The Italian participation in
the robot correctly copes with the changing environ- RoboCup was boosted by the creation of a national team,
ment; it must be fail safe with graceful degradation in called ART (Azzurra Robot Team) [21], formed by several
order to ensure that the robot may operate also in case universities and the Consorzio Padova Ricerche. ART ob-
of damages; the hardware system of the robot must tained the 2nd place in 1999 and subsequently it was split
be low power designed to optimize the batteries, and into several local teams: Golem, Artisti Veneti and Milan
so on. From this point of view, several of the typical RoboCup team.
challenges of embedded systems are also challenges The Four Legged Robot league is played in 4x6 meters
for robotics systems. field by 4 four-legged Aibo robots. The Aibo have on
board a color camera and their mechanical structure pro-
Human Robot Interface The field of Human Robot In- vides 18 degrees of freedom. The availability of a standard
terface (HRI) is related to the interaction modalities platform has significantly contributed to the scientific eval-
between the user and the robot. This field may be sub- uation of the solutions proposed. The SPQR team partici-
divided into two subfields: the cognitive HRI (cHRI) pated in the competition since 2000 obtaining the 4th place
and the physical HRI (pHRI) [2]. Cognitive HRI an- and accessing the quarter finals several times.
alyzes the flow of information between the user and
the robot and it mainly focuses on interaction modal- Recently, a Humanoid Robot league started to approach
ities, which may span from textual interfaces to voice the ultimate goal of RoboCup to build a humanoid team to
and gestures. The interface may be more or less intel- play with humans [17]. Humanoid Robotics is currently
ligent in the sense that the robot may be constrained one of the main challenges for many researchers, mostly
by a fixed set of commands or it may interpret a string focussing on mechanics and locomotion. Politecnico of
written in natural language or a sequence of gestures Torino developed the humanoid robot Isaac that has partic-
performed by the operator. The interface may also ipated to RoboCup Humanoid League since 2003. IAS-
be adaptive in the sense that the robot may adapt to Lab of University of Padova later joined the Humanoid
the operator through a suitable training phase. Phys- League, with a fully autonomous humanoid robot that uses
ical HRI instead concerns the design of intrinsically an omnidirectional visor.
safe robots. The main idea is to interpose compli- It is worth emphasizing that the ART national model led
ant elements between motors and moving parts of the to scientific and technical success: ART showed the abil-
robot in order to prevent damages in case of impact, ity to realize competitive robotic football players, but fore-
and without performance loss. Hence, cHRI research most the ability to blend in a single national team method-
is closely related to the research of AI and Robotics, ologies and implementation techniques individually devel-
while pHRI research is more linked with research in oped by the research groups. In this respect, the work
Industrial Robotics. done on the issue of coordination, leading to the definition
of communication and coordination protocols used by the
ART players [16], has been both very challenging and very
5 Applications successful. Finally, collaboration/competition achieved in
the project has been essential to the final results, since
In this section, we report on a few application scenarios,
it allowed for a project development with a tight interac-
where the research on Artificial Intelligence and Robotics
tion and exchange of results, compared to conventional re-
has been developed in Italy.
search projects.

5.1 Robotic Soccer

RoboCup started its activity about ten years ago by tak- 5.2 Rescue Robotics
ing soccer games (football for Europeans), as a scientific
testbed for the research in AI and Robotics. Italian re- Besides soccer, RoboCup promotes other leagues, aiming
searchers gave a significant contribution to RoboCup over at the transfer of the research results into socially and in-
the years, both at the organization level and in terms of par- dustrially relevant contexts. Specifically, RoboCup Res-
ticipating teams. RoboCup 2003 was held in Padova [24], cue [18] aims at the design of systems to search and res-
and it attracted more than a thousand participants from all cue for large scale disasters. Here we focus on the res-
over the world. Below we focus on the leagues, where the cue robot league, that aims at the design of robots search-
Italian participation has been more relevant. ing victims in an unknown environment representing a dis-
The Middle-Size league is played within a 5x9 meters aster scenario. This kind of application brings in scien-
field by 4 wheeled robots per team and the body of the tific challenges, related to the uncertainty about the envi-
robot must be within a cylinder of 50 cm diameter and 80 ronment, that are not present in the soccer leagues. The
cm height. All sensing devices must be onboard the robots, experimental set up, called arena, is being developed in

Anno III, N° 1/2, Marzo-Giugno 2006 91

 CONTRIBUTI SCIENTIFICI

close cooperation with USAR 2 . The arenas have already 5.4 Robotics for Elderly and Impaired People
been used in various experiments (including RoboCup and
AAAI rescue competions) and nowadays represent a refer- The goal of the project RoboCare4 sponsored by Italian
ence for experimental evaluation of the performance of res- Ministry of Education, University and Research (MIUR)
cue robots. The current aim of the competition is twofold: from 2002 to 2006 is to build a multi-agent system which
mobility and autonomy. As for the former, the research generates user services for human assistance. The system
is focussed on the mechanical design that allows the robot is implemented on a distributed and heterogeneous plat-
to overcome the obstacles present in the environment; the form, consisting of a hardware and software prototype.
latter is concerned with the design of robots that can au- The project, currently running, is coordinated by the
tonomously explore the environment, possibly working in ISTC-CNR Roma, subproject units are at the Universities
a team, build the map, find the victims and locate them in of Genova, Torino, Bologna, Parma, Roma “La Sapienza”,
the map. and at the CNR research centers of Genova, Palermo, and
Two Italian teams participate in these competitions since Milano.
2004: the first one from SIED Lab, within a collaboration The use of autonomous robotics and distributed comput-
between ”Istituto Superiore Antincendi” and the Univer- ing technologies constitutes the basis for the implementa-
sity of Rome ”La Sapienza”; the second one from the AL- tion of a number of services in an environment with el-
COR lab of the University of Rome “La Sapienza”, which derly people, such as a health-care institution or a home
developed a model-based approach to the executive con- environment. The fact that robotic components, intelligent
trol of a rescue rover, winning the third award in 2004. systems and human beings are to act in a cooperative set-
The RoboCup activity contributed and benefitted from the ting is what makes the study of such a system challenging,
results of the research project Simulation and Robotics for research and also from the technology integration point
Systems for Operations in Emergency Scenarios (SRSOES of view.
2003-2005), funded by Italian MIUR 3 . The project is organized in 3 tasks: the development
of a HW/SW framework to support the system; the study
5.3 Space Robotics and implementation of a supervisor agent; realization of
robotic agents and technology integration. Alongside the
The aim of the project An Intelligent System for the Super- above research tasks, common usability and acceptability
vision of Autonomous Robots in Space, funded by the Ital- issues are analyzed, contributing to the implementation of
ian Space Agency (ASI) during years 1997-2000, is the ap- SW development, visualization and simulation tools for
plication of AI techniques to the design and realization of multi-robot systems.
space robotics systems for planetary exploration missions,
that require an increasing autonomy. In particular, the aim REFERENCES
of this project has been the application of AI techniques to
the design and realization of an effective and flexible sys- [1] R. C. Arkin. Just what is a robot architecture any-
tem for the supervision of the ASI robotic arm SPIDER. way? Turing equivalency versus organizing princi-
The project was coordinated by the unit at the Univer- ples. In AAAI Spring Symposium on Lessons Learned
sity of Palermo. Subproject units were the Universities of from Implemented Software Architectures for Physi-
Roma “La Sapienza”, Torino, Genova, Parma, and the re- cal Agents, 1995.
search centers ISTC-CNR Roma and IRST-ITC Trento.
[2] A. Bicchi, and G. Tonietti. Fast and soft arm tac-
The scientific objective of the project is the design and
tics: Dealing with the safety-performance tradeoff in
development of an intelligent system able to supervise au-
robot arms design and control. IEEE Robotics and
tonomous robots in space. The system is based on a multi-
Automation Magazine 11(2), 2004.
agent architecture in which each block is a software agent
interfaced with the rest of the system. This design choice is [3] A. Bonarini, M. Matteucci, and M. Restelli. Filling
motivated by high flexibility, agent interchangeability with the gap among coordination, planning, and reaction
consequent easy improvement of the architecture, reuse of using a fuzzy cognitive model. In RoboCup 2003:
all the agents or part of them, or of the architecture itself. Robot Soccer World Cup VII, pages 662–669, Berlin,
The architecture has been designed by keeping in mind the Heidelberg, 2003. Springer-Verlag.
ASI missions; but it is fully general and the single mod-
ules and the whole architecture may be easily reconfigured [4] R. A. Brooks. A robust layered control system for a
for the supervision of other robotic systems. The project mobile robot. IEEE Journal of Robotics and Automa-
aimed at realizing an innovative research product, and it is tion, 2(1), 1986.
complementary to ASI activities.
[5] C. Castelpietra, A. Guidotti, L. Iocchi, D. Nardi,
2 Performance Metrics and Test Arenas for Autonomous Mobile and R. Rosati. Design and implementation of cog-
Robots. www.isd.mel.nist.gov/projects/USAR/
3 sied.dis.uniroma1.it 4 robocare.istc.cnr.it

92 Anno III, N° 1/2, Marzo-Giugno 2006

 CONTRIBUTI SCIENTIFICI

nitive soccer robots. In RoboCup 2001: Robot Soc- [19] Y. Lesperance, H.J. Levesque, F. Lin, D. Marcu,
cer World Cup V, pages 312–318, Berlin, Heidelberg, R. Reiter, and R. B. Scherl. A logical approach to
2002. Springer-Verlag. high-level robot programming. In AAAI FAll Sympo-
sium on Control of the Physical World by Intelligent
[6] A. Chella, M. Frixione, and S. Gaglio. Understanding
Systems, 1994.
dynamic scenes. Arti cial Intelligence, 123:89–132,
2000. [20] H.-H. Nagel. Steps toward a Cognitive Vision Sys-
tem. AI Magazine, 25(2):31–50, 2004.
[7] A. Chella, S. Gaglio, and R. Pirrone. Conceptual
representations of actions for autonomous robots. [21] D. Nardi et al. ART-99: Azzurra Robot Team. In
Robotics and Autonomous Systems, 34:251–263, RoboCup-99: Robot Soccer World Cup III, pages
2001. 695–698. Berlin, Heidelberg, 1999. Springer-Verlag.
[8] L. Chittaro and A. Montanari. Efficient temporal rea- [22] N.J. Nilsson. Shakey the robot. Technical Report
soning in the cached event calculus. Computational 323, SRI International, Menlo Park, CA, 1984.
Intelligence Journal, 12(3):359–382, 1996.
[23] S. Nolfi and D. Floreano. Evolutionary Robotics.
[9] S. Coradeschi and A. Saffiotti. An introduction to MIT Press, Cambridge, MA, 2000.
the anchoring problem. Robotics and Autonomous
Systems, 43(2-3):85–96, 2003. [24] E. Pagello et al. RoboCup-2003: New Scientific and
Technical Advances. AI Magazine, 25:81–98, 2004.
[10] P. I. Corke. Visual Control of Robots: High-
Performance Visual Servoing. Wiley, New York, [25] R. Pfeifer and C. Scheier. Understanding Intelli-
1996. gence. MIT Press, Cambridge, MA, 1999.
[11] G. De Giacomo, L. Iocchi, D. Nardi, and R. Rosati. [26] M. Piaggio. Classifying robot software architecture.
A theory and implementation of cognitive mobile AI*IA Notizie, 4, 1998.
robots. Journal of Logic and Computation, 5(9):759–
785, 1999. [27] R. Reiter. Knowledge in action: Logical foundations
for describing and implementing dynamical systems.
[12] R. Fikes and N. Nilsson. STRIPS: A new approach to MIT Press, Cambridge, MA, 2001.
the application of theorem proving to problem solv-
ing. Arti cial Intelligence, 2, 1971. [28] R. Reiter and A. Mackworth. A logical framework
for depiction and image interpretation. Arti cial In-
[13] A. Finzi and F. Pirri. Combining probabilities, fail-
telligence, 41:125–155, 1989.
ures and safety in robot control. In Proceedings of
IJCAI-01, pages 1331–1336, 2001. [29] S.J. Russell and P. Norvig. Arti cial Intelligence: A
Modern Approach. Pearson Education, 2003.
[14] E. Giunchiglia, G. N. Kartha, and V. Lifschitz. Repre-
senting action: Indeterminacy and ramifications. Ar- [30] L. Sciavicco and B. Siciliano. Modelling and Con-
ti cial Intelligence, 95(2):409–438, 1997. trol of Robot Manipulators, 2nd ed. Springer-Verlag,
[15] J. Hallam and H. Bruyninckx. An ontology of Berlin Heidelberg, 2000.
robotics science. In H.I. Christensen, editor, Euro- [31] M. Shanahan. Perception as abduction: Turning sen-
pean Robotics Symposium 2006, pages 1–14, Berlin, sor data into meaningful representation. Cognitive
Heidelberg, 2006. Springer-Verlag. Science, 29:103–134, 2005.
[16] L. Iocchi, D. Nardi, M. Piaggio, and A. Sgor-
[32] L. Steels. Towards a theory of emergent functional-
bissa. Distributed coordination in heterogeneous
ity. In J.A. Meyer and S.W. Wilson, editors, From
multi-robot systems. Autonomous Robots, 15:155–
Animals to Animats I, Cambridge, MA, 1991. MIT
168, 2003.
Press.
[17] H. Kitano and M. Asada. Robocup humanoid chal-
[33] M. Veloso, W. Uther, M. Fujita, M. Asada, and H. Ki-
lenge: That’s one small step for a robot, one giant
tano. Playing soccer with legged robots. In Pro-
leap for mankind. In Proc. of IEEE/RSJ International
ceedings of IROS-98, Intelligent Robots and Systems
Conference on Intelligent Robots and Systems 1998
Conference, Victoria, Canada, October 1998.
(IROS 98), pages 419–424, 1998.
[18] H. Kitano et al. Robocup-rescue: Search and res-
cue for large scale disasters as a domain for multi-
agent research. In Proceedings of IEEE Conference
on Man, Systems, and Cybernetics(SMC-99), 1999.

Anno III, N° 1/2, Marzo-Giugno 2006 93