Catálogo de publicaciones - libros

Due to the evolutionary nature of humans, systems are required to be continuously replaced because of their inability to meet or adapt to the changing needs of humans. Such replacements are costly and time consuming. Therefore this paper proposes a cheaper and quicker alternative, which is to embed self-evolving mechanisms into a multi-agent environment, thus giving the agents in the environment the ability to either meet or exceed the way in which humans evolve.

To survive the increasing worldwide competition, new technologies are required by product manufacturers to improve both effectiveness and efficiency of collaborative design and. In this paper, a computational model of collaborative product design is proposed to facilitate the management and coordination of the collaborative product design process. Based on this model, the system architecture of a multi-agent framework for collaborative design is discussed. A software prototype featured by flexibility, scalability, distributed decision-making, and etc., is now being developed.

Aiming at image target recognition, a novel algorithm of Gabor wavelet neural network based on parallel structure is proposed in this paper, and the system of neural network for multi-target recognition is designed. Based on the characteristics of multi-CPU parallel structure and the parallel property of neural network, the algorithm of Gabor wavelet neural network is proved theoretically. The relevant algorithm structure is designed; the training and recognizing algorithms for image target recognition are given out. Finally, the simulation experiment for 4 types of plane targets indicated that recognition rate reached 98%+, recognizing time was 40ms.

Learning a Nash equilibrium of a game is challengeable, and the issue of learning all the Nash equilibria seems intractable. This paper investigates the effectiveness of a momentum-based approach to learning the Nash equilibria of a game. Experimental results show the proposed algorithm can learn a Nash equilibrium in each learning iteration for a normal form strategic game. By employing a deflection technique, it can learn almost all the existing Nash equilibria of a game.

The research result of neurobiology shows that emotions have a great effect on the intelligence of human beings. To make agent have the ability of handling emotions, a new model structure of emotion agent is proposed, based on the traditional BDI agent model. The model firstly establishes a new emotion knowledge base for emotion expression, secondly performs emotion reasoning by an emotional reasoning algorithm, and finally implements the emotions treatment. An emotional agent based on this model has been realized, experiment results show that it is efficient to transact simple emotions.

In order to realize collaboration and negotiation among MAs (Mobile Agents) in MAS (Mobile Agent System), or control the Mobile Agent and release the resources it occupied, There must be a reliable and efficient communication mechanism for mobile agent, it plays a very important role in the realization of MAS and in the promotion of the widely use of MAS in the large scale network environment. In this paper, we present a novel communication mechanism MRTBDCHP (Multi Region-Tree Based Dynamic Commission Home Proxy Communication Mechanism) for Mobile Agent, it can temporarily commit a server as its communication support point near the resource or its associators, it overcomes many shortcomings of the currently existing mechanisms, further more, it is more advantageous for the application of MAS in the large scale network environment, and this mechanism is more conform with the design principles of distributed computing systems.

Aimed at the cooperation-oriented software system, this paper advanced a novel model of software architecture based on cooperation. Even as the component is the aggregation entity of objects, the cooperation is defined as the aggregation of agents based on the common tasks and the environmental constraints. As to the description of cooperation-oriented architecture, this paper brought forward a solution based on the extended WRIGHT. This new language of architecture description is named as WRIGHT*.

The paper proposes an agent-based constraint programming architecture that we have successfully applied to solve large, particularly combinatorial, operations problems. The architecture is based on the asynchronous team (A-Team) in which multiple problem solving agents cooperate with each other by exchanging results to produce a set of non-dominated solutions. We extend the A-Team by introducing CSP-specific agents, explicitly defining solution states, and enabling solution decomposition/composition, and thereby improve the performance, reliability, and automation of constraint programming significantly.

To enable efficient and appropriate uses of the resources in web, it is important to provide range query to keep track of the availability and attributes of millions of service which are geographically distributed. Range query in P2P face a performance tradeoff problem between the efficiency of query and number of message: Optimizing one tends to put pressure on the others. In this paper, a range query algorithm had proposed which lookup for every node in range can be done via O(logN) hops, and number of messages is trend to O(logN)+m-1. We evaluate our system in this paper via a simulation and show that its design along with particular Range-query algorithm meets the goals of efficient query and low message load.

With the rapid development of computer technology, the distributed applications scale up increasingly. Web service becomes more useful, and more software systems begin to make use of service-oriented architecture SOA. To improve the dependability and scalability of SOA, one effective way is to provide service replicas and balance loads among the replicas via adaptive load balancing service based on the middleware. By using middleware, we can satisfy the urgent demands of performance, scalability and availability in current distributed service-oriented applications. However, most of the current load-balancing middleware adopt the per-replica load monitoring granularity, and if multiple kinds of service groups are deployed to the same host problems will arise such as redundant load monitoring and weak scalability. To solve these problems, we design and imple-ment a multiple-granularity load balancing middleware model by using agents. In this paper, we will present the architecture of our model with the simulation results.