Catálogo de publicaciones - libros

Currently industrial automation systems are built using a hierarchical top-down approach, yielding tightly coupled and low flexibility systems. Holonic and intelligent agent-based industrial control systems have the potential to be much more highly robust and flexible systems with very loose coupling between subsystems. Despite this potential these systems are slow to be adopted by industry. This paper explores Rockwell Automation’s current agent philosophy, application experience, the obstacles to widespread adoption of agent technology in industrial automation systems, and its recent activities to overcome some of the obstacles.

This paper goes back to the origins of the Holonic Systems concept, a wording coined by Arthur Köstler [1] but actually based on fundamental insights from Nobel Prize winner Herbert Simon [2]. Simon’s theme is limited rationality and its implications for the ability to create and sustain sophisticated artifacts in the dynamic and demanding environments that are characteristic for today’s society. Holonic and multi-agent systems are amongst the most complex artifacts emanating from deliberate human design and development activities. Therefore, this paper presents these fundamental insights from Simon, augmented with more recent research results on complex adaptive systems [3], and discusses implications for the design of Holonic Multi-Agent Systems. In particular, the development of subsystems (holons) suited for incorporation into larger systems, at some later stage and without knowing these larger systems in much detail, is at the center of the discussions in this paper

MAS (Multi-Agent Systems) and HMS (Holonic Manufacturing Systems) are enabling the vision of the Plug & Play Factory and paving the way for future autonomous production systems. Previous research activities and pilot implementations reported the FIPA (Foundation for Intelligent Physical Agents) set of standards as a facilitator for the interoperation of heterogeneous agents in manufacturing control and planning. One of the current challenges for the design and implementation of intelligent agents is the simulation and visualization of the agent societies. This issue is significant as soon as the software agent is embedded into a mechatronic device or machine resulting in a physical intelligent agent with 3D-mechanical restrictions. This paper provides a review of the current software tools for the simulation and visualization of MAS in industrial applications and presents a software suite for the three-dimensional creation, simulation and visualization of agent societies. The documented research describes the methodology for the 3D representation of individual agents, the related identified objects present in the interaction protocols, and the assembly features and clustering algorithms. Finally, the paper describes how to capitalize the 3D information for controlling and planning of industrial assembly operations.

Developments in Holonic Manufacturing Systems (HMS) have been reported in three main areas: architectures, algorithms, and methodologies for HMS. Despite the advancements obtained in the first two areas the methodologies for HMS have not received great attention. To date, many of the developments in HMS have been conducted in an almost “empirical way”, without design methodology. There is a definite need to have methodologies for HMS that can assist the system designer at every development steps. This methodology should also provide clear and unambiguous analysis and design guidelines. To this end, in this work we present a Multi Agent Methodology for HMS analysis and design.

Multiagent systems (MAS) provide a useful approach to distributed problem solving whereby the growing complexity of robust and scalable systems increases the requirements on efficient structures and organisations. A promising approach to organisation of MAS is the theory of holonic multiagent systems (HMAS), which allows different kinds of recursive agent groupings (holons) whose members temporarily collaborate to solve a certain sub-problem. We present a holonic multiagent system based on a probabilistic agent architecture which is efficient for agent-based distributed data mining and simulation. We illustrate its ability to build different kinds of high-performance holons which are useful for different simulation aspects. The holons are built using cloning and merging of probabilistic networks representing the behaviour and knowledge of the agents. Finally, we show their benefits within the project SimMarket concerning supermarket simulation based on a framework for probabilistic HMAS.

An agent architecture is described with high-level process management applications in mind. The agent is “information-based” in that it is designed to operate with real-time information streams. The information that it receives may be of questionable integrity, and is represented in probabilistic first-order logic. The agent applies entropy-based inference techniques to its information base to determine its actions. Each agent is distinguished by the its information and by its plans. The agents’ information consists of individual chunks that can be formed into clusters and so into sub-clusters, thus permitting the design of hierarchic multiagent systems consistent with holonic principles.

A difficulty encountered when developing Holonic Manufacturing Systems (HMSs) is the need to place some or all of the intelligence associated with a holon on a small, low-powered device that has real-time constraints. This requirement must be balanced with the need to communicate with such a holon using standard, open, and sophisticated protocols. A possible solution is to split the functionality into two parts with a lightweight sub-holon on the device and an associated high-level sub-holon on a remote server. However a key difficulty is ensuring that the communication protocol between the two sub-holons is robust and reliable, while still allowing all the flexibility that is required in these types of applications. This paper explores the use of Elementary Net Systems as a basis for the specification of the communication and demonstrates that this provides a simple and robust basis for designing correct communication protocols in the case where communication is via a shared data table, such as the memory on an RFID tag.

One of advantages for software agent is their ability to serve as proxies for trade and bartering. This has led to the analysis and development of negotiation mechanism for general resource allocation problem. We consider agent-mediated Pareto optimal allocation of resources through market mechanism. We formulate virtual market model as a discrete resource allocation problem in dynamic situations, and demonstrate the applicability of the market concept with multi-agent paradigm to this framework. In this paper we clarify the proposed mechanism successfully calculates Pareto optimal solutions for the resource allocation problem by comparing our method with conventional analytic approaches. Additionally we apply the mechanism into dynamic market environment, and analyse the rationality of the emerged market by computer simulation.

Communication between agents is possible only when agents understand ontologies used in exchanged messages. Since we cannot hope that one universally accepted ontology would be ever created, approaches allowing using different ontologies by different agents must be used. In this paper, we present a way of expressing mappings between ontologies and using these mappings for communication between agents. To illustrate our approach, we have chosen several ontologies that describe transportation domain and created partial mappings between them. The mappings are expressed as matching forward rules which can be applied to messages exchanged between agents. In this way, messages are translated between different ontologies, as we demonstrate on an implementation that uses Jade and Jena packages. Using this approach, agents from different departments within one company or from different companies can be integrated even when they use different ontologies.

Classical control systems are based on feedback techniques and models that generally cannot manage computational complexity, nonlinearity and uncertainty. Moreover, classical control cannot adapt well to the variability of the processes under control in a dynamic fashion. However, agent-based control eases combinatorial complexity by enabling a robust partitioning of knowledge and behaviors. It is a difficult challenge to create the infrastructure, development system and validation tools for agent systems. In this paper we discuss fundamental steps to achieve the foundation infrastructure for creating agents but also we address several guidelines to create the agents and the requirements to present this to non-agent specialists.