Catálogo de publicaciones - libros

This paper describes a hybrid dynamical system-based approach to formalizing and mechanizing analyses of dynamical agents, i.e., situated, embodied actors that continuously respond to their environment. As an example, the paper describes a class of formalized metrics for reasoning about the relative difficulties of agent navigation in various environments —not just whether one scenario is more difficult than another, but how much more difficult a scenario might be— and presents results of relative difficulty reasoning using a specific example metric. This illustrates that qualitative or heuristic agent properties, which are commonly unformalized and imprecise, may be formalized and rigorously analyzed using this approach. The paper also discusses the potential implementation of relative difficulty metrics in meta-intelligent agents.

This paper introduces a novel framework for designing multi-agent systems, called “Distributed Agent Evolution with Dynamic Adaptation to Local Unexpected Scenarios” (DAEDALUS). Traditional approaches to designing multi-agent systems are offline (in simulation), and assume the presence of a global observer. In the online (real world), there may be no global observer, performance feedback may be delayed or perturbed by noise, agents may only interact with their local neighbors, and only a subset of agents may experience any form of performance feedback. Under these circumstances, it is much more difficult to design multi-agent systems. DAEDALUS is designed to address these issues, by mimicking more closely the actual dynamics of populations of agents moving and interacting in a task environment. We use two case studies to illustrate the feasibility of this approach.

In this paper, we present our approach to flexibly modelling user processes: lightweight run time agents. These agents can be thought of as essentially being ‘session handlers’ with persistent memory, which steer the execution of generic global services according to the users’ profile and its current context and goals. We illustrate this paradigm in two application scenarios, the Online Conference Service, and MaTRICS, a remote configuration management environment.

In this paper, we describe an on-going project called Concept and Sensor Networks (CSN). The development of this project has been described and discussed in past PSMP workshops [1]. The purpose of the project is to develop a framework for entities that can process sensor information into concepts. One of the features of this proposed network is the ability for the entities to use language communication to exchange concepts. These entities can potentially represent concepts, knowledge and information using different kinds of semantics. To further this project, we will to implement the proposed framework using physical and virtual sensors. In this paper, we overview the key components of the project, primarily focusing on lexical acquisition and the corresponding algorithm.

A low cost, simple, agent based, robotic sensor web to collect exploration data is a major and efficient asset to any exploration initiative. The Hybrid Analog-Digital Robotic Sensor Web (HADRS) provides this asset. HADRS is a meta-system rather than a system because it is an operational concept. This is the case, as HADRS can be used on any robotic system, of any physical configuration. HADRS will also operate in any environment, terrestrial, planetary, and space... HADRS will disperse a large number of inexpensive Mobil Agents or Mobile Units (M/U) over a wide area. Their initial search patterns will be random. The cost is low due to an inexpensive analog subsystem. When the payload instruments detect a target of interest, an autonomous, simple onboard feedback loop between the instrument and the hybrid analog/digital steering system, will guide the robot to the target.

This paper presents an AI architecture that has been developed specifically for controlling complex multi-agents interaction in games. The model is based on previous research into Emotional Societies and presents a realistic and believable environment for games. In response to a perceived lack of depth and realism in the team relationship dynamics of modern gaming, we developed a human agent architecture, multi-agent system, and demonstrative game application. The agent architecture was based partially on research into social psychology, and utilized emotion and belief representations to drive action selection. Agent interaction and relationship development was produced on the basis of the Iterated Prisoner’s Dilemma (IPD), through which a team’s success came to be determined by its members’ choices to cooperate or compete with its leader. A produced game application illustrated the operation of the developed architecture within the context of a political street protest. A set of evaluation scenarios were devised to test the success of the project work within this game application, and ultimately found it to be successful in achieving a good level of realistic team-based reasoning and interaction. Beside the potential application of the model and architecture to a computer entertainment environment, the model is generic and can be used as well for “serious” application which involves distributed emerging behavior, scenarios based simulation, complex agent-based modeling including emotional, reactive and deliberative reasoning.

In this article, we intend to characterize, at least on BDI (Belief-Desire-Intention) basis, a class of games G for which there is a MAS and vice-versa. As a consequence, criteria of rationality for agents can be directly applied to players and vice-versa. Game analysis formal tools can be applied to MAS as well. The main results of this article are the following two lemmata.

Lemma I: Every MAS belonging to G is, essentially, a Game.

Lemma II: Every Game can be implemented as a MAS and Equilibria are Optimal Desires Satisfaction.

As the information technology rapidly expands (regardless of the technology area) collaboration, information sharing and distributed computing become inevitable phenomena (i.e. e-commerce, e-health, e-science etc.) [1]. Since the complexity is a driving force behind the enhancements, new network-centric infrastructures and solutions are proposed to address the problem. Grid is an example solution that handles the situation in a cost-effective and efficient manner. There are available grid based infrastructures to solve scientific (or other) type of applications which need harmony among distributed hardware, software resources and human users. To achieve better usage of resources in a secure and intelligent way, scalable, reliable, and dynamic infrastructures are current requirements. Based on the current trends, we propose a secure, dynamic, and intelligent resource management and access method employing mobile agents along with the grid infrastructure.

In this paper we provide a formal definition of the component concept using the Heterogeneous Flow Systems Specification formalism (HFFS) a general system theory approach to the representation of timed systems. The HFFS formalism permits the development of hierarchical and modular components whose structure evolves over time. HFSS mobile components, a particular form of structural evolution, can represent mobile entities that exist in many systems like, for example, mobile communication devices and mobile software agents. The formal description is used to model a simple control application, based on mobile software agents, where a mobile hybrid component acts as a digital controller, allowing the runtime update of the control system of a vehicle.

Mobile agents were introduced as a new design paradigm for distributed systems. One advantage of mobile agents is to reduce network traffic as compared to the client-server paradigm, simply by moving code close to the data instead of moving large amount of data to the client. Unfortunately, many mobile agent toolkit suffer from to simple migration techniques. Therefore, we argue in this paper for an new migration technique that supports an adaptive decision on the code and data relocation technique. We propose several techniques for code analysis and alteration and present an algorithm to determine the optimal migration strategy within this model under the assumption of full knowledge about the application and network environment.