MARCO SOMALVICO S LEGACY ON MULTIAGENT SYSTEMS

Transcription

1 MARCO SOMALVICO S LEGACY ON MULTIAGENT SYSTEMS Francesco Amigoni, Nicola Gatti, Viola Schiaffonati Artificial Intelligence and Robotics Laboratory Dipartimento di Elettronica e Informazione Politecnico di Milano Piazza Leonardo da Vinci 32, Milano, Italy Abstract. In this paper we present Marco Somalvico s and our research on multiagent systems through the parallelism with the three parts of the human cerebral cortex devoted to self-preservation, to emotions and feelings, and to the activities related to the use of symbolic language. The proposed taxonomy, firstly inspired by Marco himself, reflects his tendency to abstraction and to bridging together different disciplines. 1. INTRODUCTION According to Marco Somalvico s innate ability to find fruitful connections between different disciplines, in this paper we discuss the researches he stimulated about multiagent systems in the light of the metaphor of the three units of the human brain [22] (see also the brief summary in [21]). Throughout the biological human evolution, the cerebral cortex has developed three components that progressively appeared and became superposed: the oldest (archipallium), located underneath and to the back; the next one (paleopallium), resting on an intermediate position, and the most recent (neopallium), situated in the top front. These three cerebral layers, although interconnected, provide peculiar functions. 1 - The archipallium, or primitive brain, is responsible for self-preservation. Mechanisms of aggressions, repetitive behaviors, instinctive reactions, involuntary actions, and visceral functions (cardiac, pulmonary, intestinal, etc.) occur in this cerebral unit, which also controls other functions, like approach, attack, flight, mating, and ritualism. 2 - The paleopallium, or intermediate brain, comprises the structures of the limbic system that are connected with the emergence of the inferior (primitive) mammals. This system commands certain behaviors that are necessary for the survival of all mammals: the ability to distinguish between the agreeable and the disagreeable, specific affective functions (such as nursing and protecting toddlers), ludic behaviors, and the emotions and feelings. 3 - The neopallium, or superior (rational) brain, is typical of superior mammals. It is related to symbolic language (and thus to skillful intellectual tasks such as reading, writing, and performing mathematical calculations). It has been called by Paul MacLean the mother of invention and the father of abstractive thought.

2 Since our researches on multiagent systems currently ongoing at the Artificial Intelligence and Robotics Laboratory are largely heterogeneous, we present them as related to the activities of these three cerebral units, according to an idea of Marco himself. First, we address the multiagent topics that can be traced down to some of the functions of archipallium and paleopallium; then we consider the topics that can be related to the functions of neopallium. We consider only multiagent systems in which benevolent agents cooperate toward the achievement of global goals which we call, following one of the Marco s eccentricities, agencies to stress their nature of unitary machines [18]. The name agency in the sense used here was firstly introduced by Marvin Minsky [25] under the metaphor of the society of mind and has been also used in [23, 24]. 2. AGENCIES, THE ARCHIPALLIUM, AND THE PALEOPALLIUM The research we have performed about multiagent systems, or agencies, has been driven from the beginning by the architecture and methodology of dynamic agency. This paradigm aims to identify and experimentally validate a flexible methodology for the development of cooperative, software and robotic, multiagent systems. The dynamic agency architecture conceives each agent as composed of a pair of semiagents: the operative one offers specialized abilities to operate in an environment, while the cooperative one integrates the agent in a uniform cooperation framework. The dynamic agency methodology prescribes the development of an agency as articulated in a sequence of phases that involve, firstly, the identification of the most adequate operative semiagents for the task at hand [19, 20] and, secondly, the installation of the cooperative semiagents by possibly adopting the modern powerful software technique of mobile code systems [15, 16]. The flexibility exhibited by this paradigm in the management of the cooperative semiagents facilitates the reuse of the agents and the dynamic variation of the agency composition and structure ANTHROPIC AGENCY (SELF-PRESERVATION) This research aims at developing applications of multiagent systems to guarantee, in a general sense, the humankind preservation. In this context multiagent systems can be adopted to address two aspects: the modeling and regulation of the human body processes and the provision of aids and instruments to address different impairments. Concerning the first aspect, the adoption of the multiagent paradigm enables distributed regulation systems that are suitable for complex phenomena and that are adaptable at several levels [3]. They allow the parameter tuning (and learning) for single regulation models, for the model fusion techniques, and for the model network architecture and composition. The multiagent architecture we propose - called anthropic agency - is based on a cooperative interaction among agents [4] and has been employed for the regulation of the glucoseinsulin metabolism [2] and for the regulation of the pace-makers frequency rate. Concerning the second aspect, the reduction of the size and energetic consumption of the electronic devices envisages an increasingly wider dissemination of hardware and software agents working on (and, sometimes, inside) the human body. In this way, it is

3 possible to equip the human body with devices for biomedical purposes (such as prostheses or safeguard systems) that can cooperate with other similar devices located in appropriately equipped environments. We are working in particular on some projects oriented to build such anthropic and environmental devices for disabled people. These projects address vocal impaired people (Bliss project), tetraplegic people (Poli-TAM project), and Down s syndrome children (Amico project) MINERVA AGENCY (LUDIC BEHAVIORS AND TEACHING ACTIVITIES) In this research we have developed a system, called Minerva, able to automatically organize virtual museums, given the set of works of art and the environments in which they must be displayed, and to support the user in the visit and fruition of these museums [10]. Both activities can be metaphorically related to the ludic behaviors of the paleopallium. Minerva blends together the classical inferential techniques of artificial intelligence and the paradigms and technologies of multiagent systems: it includes agents and components that are modular and independent in their activation and that communicate and cooperate in achieving the museum organization. Minerva has been experimentally implemented in two different versions: the first one arranges archeological finds for their allocation in historical buildings; the second one arranges design objects for their allocation in imaginary virtual environments. Furthermore, we have started to apply the same kernel of Minerva for the management of domotic devices in private houses and for the structuring of electronic courses, on the basis of the abstract idea that in both cases an organization of objects in some environments (both real and virtual ones) is involved ROBOTIC AGENCY FOR EXPLORATION AND OPERATION (TERRITORY DEFENSE) This research is metaphorically related to the territory defense function in animals, since it is about exploring and operating in a physical environment. More precisely, we have developed, according to the dynamic agency approach, an agency composed of three mobile robots and a fixed computer [15] that can tackle two different classes of tasks: the building of the geometrical map of an unknown environment [16] and the cleaning of an environment whose map is known. The work so far has been oriented to the coordination and the cooperation problems and to the possibility of dynamically switching the task while the system is operating PERCEPTIVE AGENCY (TERRITORY DEFENSE) The goal of this project, that can be seen as a special significant case of the one discussed in the previous paragraph, is the conception, design, development, and experimentation of a perceptive agency composed of robots equipped with articulated perception capabilities [1]. A perceptive agency can provide, by homogenizing the heterogeneous views returned by the perceptive agents, an abstract unitary model of extremely complex phenomena [9], such as those of interest in environmental monitoring. The applications

4 include the geometric representation of the indoor parts of buildings, the electromagnetic pollution monitoring (for which we recently obtained a two-years funding from the Italian University and Research Ministry under the program APE - Agencies for Perception in Environmental Monitoring ), the air pollution monitoring, the meteorological monitoring, and the ground and water pollution monitoring [17]. 3. AGENCIES AND THE NEOPALLIUM This part of the research is metaphorically related to the superior and intellectual activities of the brain and to the use of symbolic language and abstract thinking, since we investigate the interconnections and reciprocal influences between philosophy and artificial intelligence. In particular, we focus on the role of agencies in affording some philosophical issues from both a conceptual and a practical point of view. The research is articulated in the following more specific projects. Philosophy of artificial intelligence. This research concerns the study of the foundations of a philosophy of artificial intelligence conceived as an autonomous and mature discipline at the intersection between the human sciences and the natural sciences. The study addresses the analysis of the respective contributions of the two cultures and the systematic definition of the contact points. The ontological, gnoseological, epistemological [5], and anthropological [7] aspects artificial intelligence are also considered. Moreover, we have defined the bipolar man paradigm, conceiving a man as a unique subject who can perform his intellectual and interaction [13] activities in two different poles, called man-body pole and man-machine pole [14]. Epistemology of creativity. This research considers the conceptual paradigm of multiagent systems and agencies as a synthesis for modeling some creative activities. This synthesis conciliates two opposite approaches in the analysis of creativity: the approach which conceives creativity as connected to rationality and that which conceives it as anchored to irrationality. To join together these two traditional conceptions we have delineated a partial description of creative processes by means of agencies and we have emphasized the description of the products of such process [11]. In particular, an agency can play two roles within this scenario. It can provide an operational description of the conceptual results of the creative activity, being oriented to interpret the phenomena of reality by capturing and representing them. It can provide also a model for describing, at least partially, the creativity process itself, affording thus an operational approach to the concept of creativity [8]. Agencies and scientific research. As related to the epistemology of creativity, we have focused also on scientific discovery viewed as a peculiar case of creativity. For this reason it can be considered in the light of the agency paradigm in accordance with a modern conception of scientific discovery as a process that can be interpreted and supported by the techniques and the systems of artificial intelligence. In particular, an agency can play two roles within scientific discovery: as a support for the scientists (assistant agency) and as an operational description of the results produced by scientific activity (representational agency). We have addressed some concrete examples of modern scientific practice, taken

5 from robotics [6], medicine, and biology [12], to put special emphasis on computational and social aspects. 4. CONCLUSIONS In this paper we have presented our researches, started with Marco, on multiagent systems (called agencies in this specific context) inserted in the framework derived from the three different units of the human brain. This is fully in accordance to Marco s taxonomical tendency that has driven his visionary approach to scientific research. REFERENCES 1. Amigoni, F.; Brandolini, A.; D Antona, G.; Ottoboni, R.; Somalvico, M. (2002). Artificial Intelligence in Science of Measurements and the Evolution of the Measurements Instruments: A Perspective Conception. Proceedings of the 2002 IEEE International Symposium on Virtual and Intelligent Measurement Systems (VIMS2002), Alyeska Resort, Alaska USA, May , p Amigoni, F.; Dini, M.; Gatti, N.; Somalvico, M. (2003). Anthropic Agency: A Multiagent System for Physiological Processes. Artificial Intelligence in Medicine, Elsevier, 27(3), March 2003, p Amigoni, F.; Gatti, N. (2003). On the Simulation of Multiagent-Based Regulators for Physiological Processes. In Sichman, J. S.; Bousquet, F.; Davidsson, P. (editors); Multi-Agent-Based Simulation II (Third International Workshop MABS 2002, Bologna, Italy, July 2002, Revised Papers), Lecture Notes in Computer Science, 2581, Springer, February 2003, p Amigoni, F.; Gatti, N.; Somalvico, M. (2002). A Multiagent Interaction Paradigm for Physiological Process Control. Proceedings of the First International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS2002) Part 1, ACM Press, Bologna, Italy, July , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (1999). Processing and Interaction in Robotics. Sensors and Actuators A: Physical, Elsevier, 72(1), January 1999, p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (1999). Dynamic Agencies and Creative Scientific Discovery. Proceedings of the AISB'99 Symposium on Artificial Intelligence and Scientific Creativity, The Society for the Study of Artificial Intelligence and Simulation of Behaviour, Edinburgh, UK, April , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (1999). Some Ethical Aspects of Agency Machines based on Artificial Intelligence. Proceedings of the Fourth ETHICOMP International Conference on Social and Ethical Impacts of Information and Communication Technologies, ISBN and Abstract Book (p. 1-2), Luiss Guido Carli, Rome, Italy, October Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2000). A Multilevel Architecture of Creative Dynamic Agency. Foundations of Science, Kluwer Academic Publisher, 5(2), 2000, p

6 9. Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2001). Multiparadigmatic Cooperating Sensory Agents for Modelling Complex Systems. Proceedings of the First International Workshop on Advanced Environmental Sensing and Monitoring Technologies, Società Editrice Esculapio, Politecnico di Milano, Milan, Italy, June , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2001). Minerva: An Artificial Intelligent System for Composition of Museums. Proceedings of the International Cultural Heritage Informatics Meeting (ICHIM01) Volume 2 (Short Papers/Posters and Demos), Politecnico di Milano, Milan, Italy, September , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2001). Dynamic Agency: Models for Creative Production and Technology Applications. In Riva, G.; Davide, F. (editors); Communications Through Virtual Technology: Identity, Community, and Technology in Internet Age, IOS Press, 2001, p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2002). Multiagent Systems for Supporting and Representing Social Creativity in Science. Proceedings of the AISB'02 Symposium on Artificial Intelligence and Creativity in Arts and Science, The Society for the Study of Artificial Intelligence and Simulation of Behaviour, London, UK, April , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2002). A Theoretical Approach to Human-Robot Interaction Based on the Bipolar Man Framework. Proceedings of the Eleventh IEEE International Workshop on Robot and Human Interactive Communication 2002 (ROMAN2002), IEEE Press, Berlin, Germany, September , p Amigoni, F.; Schiaffonati, V.; Somalvico, M. (2003). The Bipolar Man Framework for Human-Centred Intelligent Systems. International Journal of Knowledge-Based Intelligent Engineering Systems, KES International, 7(1), January 2003, p Amigoni, F.; Somalvico, M. (1998). Dynamic Agencies and Multi-Robot Systems. Proceedings of the 4th International Symposium on Distributed Autonomous Robotic Systems. Karlsruhe, Germany, May In Lueth, T.; Dillmann, R.; Dario, P.; Worn, H. (editors). Distributed Autonomous Robotic Systems 3, Springer-Verlag, Berlin Heidelberg, Germany, 1998, p Amigoni, F.; Somalvico, M. (2002). Application of Mobile Code to Development of Cooperative Multirobot Systems. Proceedings of the 7th International Conference on Intelligent Autonomous Systems (IAS-7). Marina del Rey, California USA, March In Gini, M.; Shen, W. M.; Torras, C.; Yuasa, H. (editors). Intelligent Autonomous Systems 7, IOS Press, Amsterdam, The Netherlands, 2002, p Amigoni, F.; Somalvico, M. (2003). Multiagent Systems for Environmental Perception. Proceedings of the Third AMS (American Meteorological Society) Conference on Artificial Intelligence Applications to Environmental Science, Preprints CD-ROM and Abstract Volume (p. 487), AMS, Long Beach, USA, February

7 18. Amigoni, F.; Somalvico, M.; Zanisi, D. (1999). A Theoretical Framework for the Conception of Agency. International Journal of Intelligent Systems, John Wiley & Sons, 14(5), May 1999, p Amigoni, F.; Villa, M. (1999). An Algebraic Description of Agency Design. Proceedings of the ESSLLI99 Workshop on Foundations and Applications of Collective Agent Based Systems, University of Utrecht, Utrecht, The Netherlands, August Amigoni, F.; Villa, M. (2000). An Algorithm for Recruitment of Agents in Agency Design. In Lamma, E.; Mello, P. (editors); AI*IA99: Advances in Artificial Intelligence, Lecture Notes in Artificial Intelligence, 1792, Springer, April 2000, p do Amaral, J. R.; de Oliveira, J. M. (2003). Limbic System: The Center of Emotions. last accessed June MacLean, P. (1990). The Triune Brain in Evolution. Plenum Press. 23. Matsuyama, T. (1999). Cooperative Distributed Vision: Dynamic Integration of Visual Perception, Action, and Communication. In Burgard, W.; Christaller, T.; Cremers, A. B. (editors); Proceedings of the Annual German Conference on Advances in Artificial Intelligence (KI-99) LNAI 1701, Springer, p McKee, G.; Bull, J.; Schenker, P. (2002). SHAPE: Simulated Heterogeneous Agent Planing of an Environment. In Gini, M.; Shen, W. M.; Torras, C.; Yuasa, H. (editors); Intelligent Autonomous Systems 7, IOS Press, p Minsky, M. (1985). The Society of Mind. Simon & Schuster.