Catálogo de publicaciones - libros

The individual agents that interact in a multi-agent system typically exist along a continuum ranging from heavyweight cognitive agents (often of the “BDI” type) to lightweight agents with limited individual processing (digital ants). Most systems use agents from a single position along this spectrum. We have successfully implemented several systems in which agents of very different degrees of internal sophistication interact with one another. Based on this experience, we identify several different ways in which agents of different kinds can be integrated in a single system, and offer observations and lessons from our experiences.

Holonic Multi-Agent Systems (HMAS) are a convenient way to engineer complex and open systems. HMAS are based upon self-similar entities, called holons, which define an organizational structure called holarchy. An open issue of HMAS is to give holons means of self-organization to satisfy their goals. Our works focus on modeling and engineering of complex systems using a holonic organizational approach. This paper introduces the concept of as the description of agents know-how. This concept allows the representation and reasoning about agents know-hows. Even more, it encourages a reusable modeling and provides agents with means to self-organize.

There is little or no guidance to systematically design a self-organising emergent solution that achieves the desired macroscopic behaviour. This paper describes decentralised coordination mechanisms such as gradient fields as design patterns, similar to patterns used in mainstream software engineering. As a consequence, a structured consolidation of best practice in using each coordination mechanism becomes available to guide engineers in applying them, and to directly decide which mechanisms are promising to solve a certain problem. As such, self-organising emergent solutions can be engineered more systematically, which is illustrated in a packet delivery service application.

Software today is no longer monolithic, but typically part of a system consisting of many components. As engineers are no longer in control of the entire system, novel methods are sought to design complex software systems that are built from the bottom up and are robust in a dynamically changing environment. The coordination method called stigmergy that is inspired by the collective behavior of social insects is one of the candidates to help solving this problem. In this paper we make a first step in formally understanding the essence of stigmergetic behavior by studying the famous ant foraging model of Deneubourg et al. We explore the relationship between the initial (dis)order in the environment and the performance of the ant foraging behavior. We further study how this configuration of the task to solve governs the behavior of the ant colony, with special focus on the level of coordination that is achieved.

Hierarchical clustering is used widely to organize data and search for patterns. Previous algorithms assume that the body of data being clustered is fixed while the algorithm runs, and use centralized data representations that make it difficult to scale the process by distributing it across multiple processors. Self-Organizing Data and Search (SODAS) inspired by the decentralized algorithms that ants use to sort their nests, relaxes these constraints. SODAS can maintain a hierarchical structure over a continuously changing collection of leaves, requiring only local computations at the nodes of the hierarchy and thus permitting the system to scale arbitrarily by distributing nodes (and their processing) across multiple computers.

In this paper we present the Behaviosite paradigm, a new approach to affecting the behavior of distributed agents in a multiagent system, which is inspired by biological parasites with behavior manipulation properties. Behaviosites are special kinds of agents that “infect” a system composed of agents operating in that environment. The behaviosites facilitate behavioral changes in agents to achieve altered, potentially improved, performance of the overall system. Behaviosites need to be designed so that they are intimately familiar with the internal workings of the environment and of the agents operating within it, and behaviosites apply this knowledge for their manipulation, using various infection and manipulation strategies.

To demonstrate and test this paradigm, we implemented a version of the El Farol problem, where agents want to go to a bar of limited capacity, and cannot use communication to coordinate their activity. Several solutions to this problem exist, but most yield near-zero utility for the agents. We added behaviosites to the El Farol problem, which manipulate the decision making process of some of the agents by making them believe that bar capacity is lower than it really is. We show that behaviosites overcome the learning ability of the agents, and increase social utility and social fairness significantly, with little actual damage to the overall system, and none to the agents.

Modular robots represent a perfect application scenario for multiagent coordination. The autonomous modules composing the robot must coordinate their respective activities to enforce a specific global shape or a coherent motion gait. Here we show how the TOTA (“Tuples On The Air”) middleware can be effectively exploited to support agents’ coordination in this context. The key idea in TOTA is to rely on spatially distributed tuples, spread across the robot, to guide the agents’ activities in moving and reshaping the robot. Three simulated examples are presented to support our claims.

We present a simple model of self-organization for sensor networks that addresses two conflicting requirements: hazard situations should be reported to a with as little delay as possible, even in highly dynamic regimes; and power consumption by individual nodes should be as low as possible and balanced. The model includes a surveillance protocol that explores correlation between source location and event types, and a variation of that adapts continuously to energy available at selected routers and to changes in topology. Simulation runs provide support to the heuristics we implemented to select routers and nodes to report events, since network longevity increases when compared to other solutions for sensor networks. The performance increase is particularly accentuated when the correlation between event types at neighboring nodes is significant.

Explicit and high semantic level communications are not always the best approaches to coordinate the global system activity in the context of Multi-Agent Systems (MAS). Insect societies take advantage of a stigmergic way to communicate which does not require centralized control, but enable insects to coordinate their complex global tasks. In this paper, we describe and motivate a new approach to elaborate Complex Exchanges between Stigmergic Negotiating Agents (CESNA), for the critical resource sharing problem. We describe a negotiating network as a generic and suitable representation of the problem, along with its implemented behaviours. We present some promising results and attempt a first interpretation of how this decentralized system leads local behaviours to a global problem solution.

The research reported in this paper is concerned with assessing the usefulness of reinforcment learning (RL) for on-line calibration of parameters in evolutionary algorithms (EA). We are running an RL procedure and the EA simultaneously and the RL is changing the EA parameters on-the-fly. We evaluate this approach experimentally on a range of fitness landscapes with varying degrees of ruggedness. The results show that EA calibrated by the RL-based approach outperforms a benchmark EA.