Catálogo de publicaciones - libros

This paper studies how to relax the restrictions on bidders in two sequential English auctions. Existing work has showed that it is optimal for the auctioneer to auction objects in the decreasing order of the dispersion of the order statistics of the surpluses. But they assume each bidder can win at most one object. However, in many cases, each bidder needs more than one object and the objects almost affect each other. Given this, the settings that have both common and private value elements by allowing each bidder winning more than one object are studied here. Firstly, the single object model is extended to two objects without the limitation that a bidder can win at most one object. Secondly, the equilibrium bidding strategies for each auction in a sequence are described, the selling prices and winner’s total expected profits of different agendas are analyzed. Thirdly, the optimal agendas are determined. Finally, we make two experiments, one validates that it may be the optimal agenda either by increasing order or decreasing order of the dispersion of the order statistics of the surpluses, the other display the changing trend of the optimal agenda.

This paper presents the intelligent game agent that gives effective intelligence to NPCs (Non Player Characters) for which intelligence did not exist. Generally, non-player characters (NPCs), or agents, such as monsters, enemy guards, or friendly wingmen, can be controlled by a finite-state machine. To overcome the shortcoming of NPC’s restricted action, we applied a LSVM (Linear Support Vector Machine) as pattern recognition for the intelligent game agent, and processed all data in XML format to handle the data efficiently. The intelligent agent is executed in the base of the game physics engine. A lot of NPCs that act in a game learn physics values that are produced in the game, and change NPS’s action intelligently. We applied two pattern recognition algorithms to estimate the algorithm’s performance through comparison. As indicated by experiments, when the M-BP has a fixed number of input layers (number of physical parameters) and output layers (impact value), it shows the best performance when the number of hidden layers is 3 and the learning count number is 30,000. The pattern recognizer that applied LSVM shows the best performance when the learning count number is 25000, and the LSVM shows better performance than the M-BP in the intelligent game agent.

Palmprint recognition received many researchers’ attention because of it’s low resolution and cheap devices. As other biometrics, algebraic feature is the prevailing method for palmprint recognition. PCA (principal component analysis) is one prevailing algebraic transformation. It has been a successful feature detection method for pattern recognition. It deals with image vector whose dimension is usually high. 2DPCA is a novel PCA method for image matrix, and it can calculate the covariance matrix more precise. In this paper we apply the new 2DPCA method to palmprint recognition, and we make an improvement in the selection of principal components. In our method we select the principal component that is better for classification. At last we do the improved 2DPCA on the row and column direction to reduce dimension in both direction. Then we apply the method to PolyU Palmprint Database. The experiment result shows that our method got more recognition rate with lower dimensions.

It is crucial for ontology engineers to compose these heterogeneous ontologies for effective and scalable interoperability among Web agents. In this paper, we propose a combination framework based on effective and scalable ontology interoperability for high-quality query answering. We also put forward a specifical process to substantiate our combination approach. The underlying theoretical and empirical analyses are exploited for ensuring effectiveness and correctness of the approach. Experimental results from our prototype system show that high-quality query results but sacrifice some of reasoning complexity.

Organizational adaptation is one of the key issues of organization and reorganization theories. As the result of environmental changes, organizations can change their internal structures or agents in the organizations can adjust their behaviors to achieve organizational objectives. In this paper, we focus on the characteristics of organizational structures and also analyze the notion of organizational stability and its changes resulted from the constraints of roles and their relations in the organizational structure. In the end, we give a formal specification of the adaptive mechanisms for organizational structures in details to meet dynamic environment requirements.

Ubiquitous media aims to provide media services anytime and anywhere. Software agents are one of the building blocks of ambient intelligence and pervasive computing. In this paper, we use a structured overlay network as the decentralized service repository system for improved efficiency. We present an agent-based services composition framework for ubiquitous media, through a combination of the respective merits of agent and service-oriented to enables dynamic and efficient delivery of media composite services. We advocate that a service-centric context promotes media applications that allow service adaptability, deal with service availability, and support on-the-fly service composition. Finally, we present an example of our system in the context of streaming video playback involving a series of transcoding services and a mobile client.

Intentions, a crucial part of the mental state of an agent, play an important role in determining the behavior of rational agents. In order to eliminate the flaws with existing logic of intention, in this paper, we address the requests for intention semantics on formal frameworks of rational agents and the problems with existing logic of intention, put forward a novel semantics for intention, called the multi-subset possible world semantics, and its application in the formalization of intention for agent. The multi-subset possible world semantics use three subsets to describe intention. It not only avoids the logical omniscience problem and other related problems (such as side-effect problem, and etc) but also overcomes the shortcomings of the true-false subset semantics and twin-subset semantics. Compared with Konolige and Pollack’s model of intention, this semantics model doesn’t lose the reasoning ability of non-equivalent intention, and what’s more, it is simpler, more natural and satisfies the K axiom and the Joint Consistency. Our framework invalidates the problematic properties with existing logic of intention. And by imposing certain constraints on the algebraic structure of the models, many desirable properties can be obtained. At last we make an analysis for the multi-subset possible world semantics. Actually the multi-subset possible world semantics provides a new method for semantic representation of non-normal modal operators. It can be used in establishing new proper agent’s logic systems.

Under the multi-agent system (MAS) circumstances, the concurrent behaviors turn out to be very important. Concurrent behaviors can be divided into irrelative and correlative, most of them are correlative and agents should be able to deduce their behaviors. This paper put forward a concurrent Agent model based on twin-subset semantic. This model aims at the characteristics of concurrent behaviors in MAS. Based on twin-subset semantic, the logical omniscience problem and other related problems (such as side-effect problem) can be avoided. This model divides the time set into macro-time and micro-time sets, and describes the concurrency in macro-time level. In this model, parallel actions in macro-time are interleaving in micro-time level. It can be applied to establish the logical foundation for the cooperation and competition based on concurrency in MAS.

Migrating workflow system is one of a practical implementation of mobile agent-based workflow system. To enable effective negotiation and cooperation between two migrating instances of a migrating workflow, communication is an important and essential means. Previous approaches, when applied in the context of mobile agent based workflow, cannot guarantee the reliability and efficiency simultaneously. In this paper, we proposed an improved communication model to solve this problem. The model implements a computation metaphor similar to the way people correspond by mail. It introduces the concept of service domains to imitate postal areas with different postal codes, and decouple mailbox with its owners to ensure reliability. In addition, the combining utilization of Address_book and friend domain list can locate a target more quickly and exactly. Original experiments show that the model not only ensures reliable communication of moving objects, but also keeps overheads at reasonably low.

This paper presents a human tracking agent that recognizes the location and motion of the human in the home. We describe the architecture of the human tracking agent, and present an image recognition algorithm to track location and motion of the human. The human tracking agent decides the human’s location, which changes in real-time, through the reletive distance of home furniture (or appliance) and human. Unlike the human’s location, because a person’s appearance (height, weight) is different for each person, a human’s motion should be recognized to be different from each other person. We converted the image (that is acquired from the network camera) into a standard image (that is defined in the human tracking agent) for recognition of multi-user’s motion. We used a LSVM(linear support vector machine) to recognize the feature patterns for human motion. In our experiment, results of motion recognition showed excellent performance accuracy of over 80%.