Collaborative intelligent systems
Distributed robots and sensors
Our goal is the study and development of Multi-agent systems (MAS) following a two-pronged methodology that stress the importance of organization at various levels of granularity.
We study and exploit coordination, interaction, and negotiation approaches based on biological, psychological, economical, and physical knowledge to develop structural frameworks for system design and for the decomposition of the whole ensemble into interactive dynamic subsystems. We also examine relevant issues and develop methods to promote self-organization and lead to emergent efficient control frameworks, considering techniques suggested by the study of social, economic, and ecological systems, such as the well-known swarm intelligence methods. In this regard, we must note that, while again recurring to examination of structural schemes suggested by applied sciences, the emphasis here is on emulating the evolution of organizational structures rather than on matching, at some level of detail, an existing organization. In addition, we will also emphasize various system-science techniques that promote organization on the basis of notions such as the cohesion of groups of agents and interacting processes.
Regardless of the particular approach being applied, it is our intent to emphasize the importance of organization to decompose the complex design and optimization problems faced when developing distributed systems into a number of smaller optimization problems, and to developing analytical methods to evaluate the effect of alternative organization and communication frameworks on system performance. Another key concept in our studies is that of behaviour, being particularly attractive as the basis for the definition of concurrent, interacting, processes that seek multiple purposive and reactive goals.
Soft Qualitative Reasoning in Distributed Environments
Soft computing techniques, in general, and fuzzy logic methods, in particular have been the object of considerable examination as the bases for the elicitation, representation, and manipulation of objects and their relations. These methodologies have also been successfully applied to the classification of objects into relational, fuzzy-set based, semantic structures from which conceptual entities may be derived and interrelated. These structures permit to search and examine databases in terms that are meaningful to their users rather than on the basis of constructs introduced to facilitate their computational representation. In this topic, our research is focus in the application of Multiagent Systems to the generation, representation, and manipulation of conceptual structures in a distributed information environment. The foundation of our approach is provided by methods that produce descriptions of complex objects in terms that are meaningful to users of information retrieval and analysis systems. Specifically, we work in the generation of Qualitative Object Description (QOD) methods that seek to describe objects in terms of “interesting" structures and relations. These descriptions are themselves structured computational objects that, employing Zadeh's terminology, are perception-oriented rather than computer oriented. These structured descriptions are also capable of being translated into text (or hypertext) versions that facilitate examination of the objects by domain experts. Furthermore, the perception-oriented descriptions permit categorization and indexing in terms that are close to the needs of those experts..
Of particular importance in this regard are machine learning and data mining procedures that operate upon perception-based descriptions of complex objects. These methods are an important element in the elicitation of new concepts (e.g., by conceptual clustering methods) and for the discovery of relations that are actually obscured by the original, computer-oriented, representation methods. Structures discovered by these methods are important elements in the extension of semantic structures (e.g., ontologies) and in the specification of new interesting features and relations (concept formation). Our ultimate goal is the development of fundamental knowledge applicable to a wide variety of applications..
Distributed Intelligent Information Systems
The knowledge and information generated and stored in distributed environments is produced and consumed by a multitude of autonomous agents each employing different approaches and criteria. As networked knowledge repositories evolve from databases conceived, developed, and presented using a homogenous format to a user community with common interests, towards adaptive collections incorporating, integrating, and displaying information items as a function of the preferences of specific users (i.e., the collaborative WWW) a number of methods have emerged in support of the interaction of distributed autonomous agents. In this context, we are working in the development of techniques that coordinate group decision making, knowledge processing, and knowledge structuring in WWW-based systems, social networks, and other distributed environments. It is our intent to emphasize the automated derivation of perception-based descriptions of computational objects and the development of associated semantic structures (e.g., ontologies), as well as the implementation of approaches to the indexing and annotation of semantic structures, and to promote the study of methods for their management (e.g., mediation, alignment, evolution). Among these activities we focus on approaches for the discovery of information on the basis of perception-based semantic structures and for the retrieval of objects employing these structures. Specific examples of techniques of interest include methods for the development of, smart cataloguing and indexing, associative pattern and cognitive memories, development and maintenance of organizational memories, intelligent portals, personalized user interfaces, structures for management of information content in peer-to-peer and business-to-business networks, etc. Strongly related to these information and knowledge representation techniques are methods that facilitate collective decision-making, including techniques to detect and resolve conflicts, identify decision alternatives, and promote collaboration between various stakeholders in social and business networks. We are similarly interested in tools to model social-network generation, dynamics, and evolution, and in the study and facilitation of communication in social networks. In this regard we are studying similar issues as they respect to the identification, generation, and evolution of new computation paradigms, such as cloud/grid and ubiquitous computing.