Catálogo de publicaciones - libros

After September 11, we remember there was a clear conflict in public opinions within western countries. While 77% of those interviewed in France opposed to military intervention against Iraq (2003.1.9 Le Figaro), 63% of the U.S. population were proud of the U.S. role in the war (2003.3.22 CBS News). Conflicts in governmental policies are common, but conflicts in public opinion between western countries have not been observed before. Though we all share information on the Web, similar conflicts arose recently in East Asia. While about 90 percent of Chinese polled blamed Japan, more than half of Japanese polled said it was hard to tell who bore responsibility (2005.8.24 Genron NPO and Peking University). According to Global Reach, the ratio of English speaking people online has decreased to 35.2% in 2004. To increase mutual understanding between different cultures and of opinions in different languages, it is essential to build a language infrastructure on top of the Internet.

Collaborative environment will provide the architecture and infrastructure that allows seamlessly integrated comprehensive support services for collaboration involving sharing data, information, knowledge, tools for manipulation, representation and visualizations. Agent Grid Intelligence Platform (AGrIP) will aid people, teams and group to collaborate with searching processes, mining processes, decision processes and action processes. AGrIP is a highly open software environment whose structure is capable of dynamical changes. It’s a loosely coupled computer network of ever expanding size and complexity. It can be viewed as a large, distributed information resource, with nodes on the network designed and implemented by different organizations and individuals with widely varying agendas. The four-layer model for AGrIP is presented, that is data resources, maulti-agent, middleware and applications.

In this presentation I will emphasize to discuss agent model, AGrIP architecture and collaborative strategies. For collaborative strategies we have proposed flexible working flow, role assignment, ontology-based collaboration, policy driven and planning.

While industrial robots typically perform pre-programmed and routine manufacturing jobs for humans, intelligent service robots serve and interact with people intelligently with its own perception and cognition of external circumstance. As an intelligent system, intelligent service robots are also characterized by the ability to evolve and learn to increase their performance over time.

Recently intelligent service robots have received great attention as one of the next-generation growth engine in many countries including Korea. In this presentation, I introduce the research activities on intelligent service robots in Korea and then provide an agent-based approach to the integration of various distributed functionalities to overcome the inherent complexity of intelligent service robots in dynamic environments.

The rapid growth of web along with the increasing decentralization of organizational structures has led to the creation of a vast interconnected network of distributed electronic information in many fields such as medical science, bioinformatics, high-energy physics etc. The data produced and the knowledge derived from it will lose value in the future if the mechanisms for sharing, integration, cataloging, searching, viewing, and retrieving are not quickly improved. Building upon techniques from both Semantic Web [1] and Grid [2] research areas, we propose the DartGrid [3][4] which exhibits a Dynamic, Adaptive, RDF-mediated and Transparent (DART) approach for building semantic grid applications.

In this paper we present a variant of grammar system from Formal Language theory, that seems to be appropriate for modelling dialogues in multi-agent systems. Previous attempts of simulating conversations in grammar systems used rewriting rules, which remained fix and unchangeable during the whole dialogue. The novelty of this approach is that taking inspiration from a Multi Agent Protocol language we define a grammatical system in which agents(grammars) behaviours are given by string process descriptions or protocols. Strings can be modified during running time, allowing to dynamically alter agent behaviour according to the environmental changes. This provides agents with a more flexible and adaptative reaction to unpredictable and changeable conversational space.

Formal description and verification of the interaction protocols between agents is a very valuable reserch in development of MAS. In the paper we have defined a calculus for description interaction protocols of MAS based on dialogue. The calculus is founded on process algebra and is designed to be independent of any particular model of rational agency. This makes the verification applicable to heterogeneous agent systems. With the state of session environment and formal semantics of calculus, we can verify some properties of session protocols, e.g. termination, deadlock. Our approach does not suffer from the semantic verification problem because the states of the session dialogue are defined in the protocol itself, and it is straightforward to verify that an agent is or not acting in accordance with the protocol and does not suffer from state-space explosion.

In order to synthesize stigmergy a model needs to be created that allows a collective of agents to achieve global results through local interactions in some environment. This locality of interactions between the agents and between the agent and the environment allows for the distribution of the entire system without any centralization. Stigmergy is found among social insects in nature. These natural systems show remarkable flexibility, robustness and self-organisation. These characteristics are sort after in modern software systems. Utilizing stigmergy in an artificial system allows agents to interact with one another and with the general topology in a non-centralized manner, thus giving rise to a collective solution when solving of certain tasks. Even though the agents are localized their interaction with the stigmergy layer allows other agents to be affected by the interactions. The methology of mimicking stigmergy into a software system will be described and a description of the model used to synthesize stigmergy will be given. The potential utilization of stigmergy by software agents to interact with each other and to solve certain tasks collectively is also demonstrated.

In this paper, we introduce a probabilistic epistemic temporal logic, called , which is a combination of temporal logic and probabilistic knowledge logic. The model checking algorithm is given. Furthermore, we present a probabilistic epistemic temporal logic, called , which generalizes -calculus by adding probabilistic knowledge modality. Similar to -calculus, is a succinct and expressive language. It is showed that temporal modalities such as “always”, “sometime” and “until”, and probabilistic knowledge modalities such as “probabilistic knowledge” and “probabilistic common knowledge” can be expressed in such a logic. is proven to be a sublogic of The model checking technique for is also studied.

Designing and building Intelligent robots involves integration of various functionalities such as manipulation, navigation, various recognitions, speech understanding and expression, reasoning, planning, and so on. Furthermore, such functional components are often distributed over several processors even in a single robotic system. Manifold functionalities and inherent complexity of robotic systems require a well-organized uniform control of the functional components so that the formidable integration of the functionalities becomes manageable. In this paper, we present an agent-based task management architecture for control of intelligent robots to simplify the integration task. The task manager works as an integration middleware and provides a consistent and unified control view for the functional components which may be distributed over a network.

Agent based systems are being used to solve problems ranging from Web search strategies to autonomous robots.In this paper we discuss a multi-agent based selfish routing problem in multi-channel wireless mesh networks. In this selfish routing, pairs of source-sink (destination) pairs are managed by different intelligent agents. All the agents are selfish; each agent’s goal is to minimize its own cost without considering the overall performance of the network. Meanwhile, the agents are rational, they always meet the embarrassment that whether they face the probability of causing interference for routing flow by a short cut or they route flow by a longer way steadily. We introduce the game theoretic tools to direct the agents to make feasible choices. We yield a sufficient condition of the existence of pure strategy Nash equilibria and proof it in this paper (we call this the ). Some simulations reveal the feasibility of our proposition.