D.1.3 NIAD&R: Plan for 2004

D.1.3 NIAD&R: Plan for 2004

Scientific Objectives

In the following sub-sections a more detailed description of the work we intend to pursue is presented.

D.1.3.1 Negotiation Models for Autonomous Agents

People involved: Eugénio Oliveira, Ana Paula Rocha, Henrique L. Cardoso, Luís Nogueira, Andreia Malucelli
Coordinator: Eugénio Oliveira

Research direction: Research in the context of this issue aims at developing an Electronic Institution for safe and trustable agent-based business operations. This includes appropriate models for both B2B and B2C Negotiation processes as well as to provide platforms, tools and frameworks enabling Agents' interaction in the context of Virtual Enterprise formation and operation processes.

Models and Protocols for Autonomous Agents' Negotiation

Research goals: To develop, implement and test models for autonomous, individually rational, agents representing either individuals or enterprises willing to reach fair agreements through flexible negotiation.

• To enhance previously developed Q-Negotiation algorithm.
• To develop further and test a specific tool we have been developing for B2B interaction.
• To test a specif tool for E-Brokering in the Insurance Domain.

RECENT WORK (2003):

• As a result of several meetings of AMEC (Agent Mediated Electronic Commerce) SIG of the AgentLink European Network of Excellence, the concept of Electronic Institution has emerged. In order to develop and implement this concept, i.e. a framework enabling secure and responsible (enterprises' delegates) agents'interaction we have now a first implemention of ForEV (Virtual Enterprises Formation framework). Besides all the facilities for registration and matching between mutually interested agents, ForEV already provides sophisticated negotiation protocols with learning capabilities - Q-Negotiation Algotithm.

  Q-Negotiation, developed in 2002 by Ana Paula Rocha and Eugénio Oliveira (referred in the 2002 report) is an advanced negotiation protocol for agents in the context of the Virtual Enterprise Formation stage, that encompasses the following characteristics:

• multi-issue evaluation function relying on the several different agents individual preferences;
• qualitative feedback during negotiation, enabling the agents keeping their own preferences private;
• continuous adaptation to the dynamics of the market by using a slightly modified reinforcement learning algorithm to find out the new current bid to submit to the market;
• Apropriate procedures for solving simple distributed constraints among items under negotiation, making the all process to converge to the mutually accepted solutions space.

• Moreover we have implemented and tested [Nog03a], [Nog03b], [Nog03c], a new model for automatic agent-based E-brokering in the important domain of insurance products. The implemented model encompasses different kinds of negotiation protocols between both the Broker agent and the Insurance companies agents as well as between the Broker agent and the personalised Customer agents. The main idea is to intelligently mediate by supplying Customers with meaningful information gathered from the available Insurance products. Also, by taking advantage of previously inferred customer preferences the Broker is able to negotiate different, although relevant offers and rank them for the customer. The Broker Agent is then able to incrementally build up different and separate Customers groups and associate to them meaningful (and useful) stereotypes. Special non-supervised learning algorithms are used. The final prototype has been developed on the top of a very flexible tool called Bee-Gent (from Toshiba).
• We came to the conclusion that, in order to enable flexible interaction between heterogeneous agents, a set of services based on ontologies should be provided. After a systematic study on Ontologies that has been done in the framework of a PhD thesis preparation (A.Malucelli), a first description about what these Ontology-services can be, has already been provided [Malu03]. We are also designing a specific scenario where agents representing car parts supplyers need an Ontology to interact. This Ontology has been specified through Rational Rose and edited through OntoEdit. Agents exchange messages in OXML format.

CURRENT AND FUTURE WORK:

• The Electronic Institution concept has been disigned to encompass all the stages of a Virtual Organization (Enterprise) Life Cycle: Announcement of needs, Formation, Operation and Dissolution. Our main concern is now to specify what a suitable and effective "contract" can be as a result of all the negotiation process.The "E-Contract" will include a set of procedures all the agents have agreed upon. This established "contract" has to be at the disposal of the Electronic Institution in order to permit this entity to verify, at the right time, all the important steps that should be checked and, if this is the case, to enforce previously agreed corrective actions or punishements.
• Following work on the ForEV platform, we have studied the state-of-the-art literature on e-contracting, specifically in what B2B contracts are concerned. We consider e-contracts as the formalization of a business agreement, which parties agree to sign because of lack of trust for the fulfillment of that agreement. The contract will be enforced by an Electronic Institution (considered as a trusted third-party) with appropriate services. In particular, the institution incorporates a set of norms and rules to be inherited by all contracts signed within that institution; it also includes facilities that assist contract construction (through templates), monitoring and enforcement. We are also studying deontic logic as a formal approach to model norms in contracts.
• We are enriching our model of an Electronic Institution, by considering an enhanced framework with reputation mechanisms, ontologies, transaction repositories, templates, and contract monitoring and enforcement services. As we aim at automating virtual enterprise contracts, we are defining requirements for a representation language that considers this complex setting. We consider including a flexible level of specification of e-contracts within an electronic institution, depending on trust and reputation mechanisms. Furthermore, we intend to study how agents and the institution itself can learn to adapt contract requirements to the reputation of the involved participants (adopting ideas from contract law and relational contract theory).
• Also, in order to make it more realistic for agents representing organisations (enterprises) to know how their internal processes (production and management processes) will interfere with the overall organisation operation, we are now looking at designing sound and correct inter-organisation network Workflows. Petri-Nets are good candidates for such representation, test and verification.
• BIAS, the agent-based system for E-Brokering we have implemented, is now being checked against a credible scenario, for automatic mediation of Insurance products offers and needs. Further developments will imply enhancement of agent negotiation strategies and automatic product offers on the Insurance companies side. We reserve our decision to go in this direction depending on the impact the prototype will have on potential interested software houses.
• In order to develop further the ontology-services agent, we envisage to integrate it with ForEV framework [Malu03]. Ontology can then be explored in the sense of helping different agents, representing different companies, to understand each other when negotiating their joint participation in a Virtual Organisation (Enterprise) through their commitment in supplying complementary parts of the final overall business process. There are both terminological and structural differences that may appear in the representation of a specific domain common knowledge. Ontology-based services agent will be able to recognize the potential semantic mismatch and provide a way of mutual understanding.

D.1.3.2 Agents' Adaptation, Learning and Emotions

Research direction:We are engaged in pursuing two separate research lines we believe will be definitely recognized as important in our group. These issues are an attempt to explore agents' advanced features:

(i) Multi-agent learning. The main goal of this research issue is to find an answer to the following question: "(How) can several different, heterogeneous, Learning Agents improve their performance by exchanging information during their own learning process?".

(ii) Emotion-based agents' architecture. Through this research issue we would like to answer another important question: "Will it be possible to escape from usual utility-based decision functions in which decision-making for autonomous agents is concerned?"

Multi-Agent Learning

Research goal:(How) can several different Learning Agents improve their performance by exchanging information during their own learning process?" that is the question.

RECENT WORK (2003):

• In order to answer that question, we have been studying the effects of exchanging information during the learning process in heterogeneous groups of agents in controled, and somewhat simplified, scenarios, with the objective of detecting possible improvements of agents' individual and global performances. We are using communication and heterogeneity to improve agents' learning performance in Environments that are: partially observable and dynamic; including several Agents dealing with similar problems although using different learning algorithms
• To achieve the above mentioned goal, and having in mind that first question, a method for information exchange during the learning process based on the exchange of advice amongst learners has been developed. The learning environment launches several learners solving different instances of the same problem. The instances of the problem may be either totally independent, or similar parts of the same global problem. Agents are learning either through Reinforced Learning (getting feedback from the environment as it is the case with Evolutionary Algorithms and Q-Learning) or supervised learning as it is the case with backpropagation. Agents may be at different learning stages: Exploration, Novice, Intermediate, Expert. [Oli03a], [Nun03a], [Nun03b], [Nun03c], [Nun03d], [Nun03e].

  As an example in one of our learning environments, agents try to control traffic lights in different crossings. Experiments with disconnected crossings have proved that advice-exchange, even in its simplest form, is effective. Current work is focused on solving the problems that appear when agents share the same environment and their actions affect others. In order to analyze the problems that emerge when several agents are acting concurrently in the same environment a set of experiments in the Pursuit Problem was made [Nun03e]. Results obtained allowed us to detect and solve several of the problems that caused disturbances in the learning process due to negative interactions between heterogeneous agents' actions or between the self-learning algorithms on one hand, and advice-exchange on the other.

  Giving advice consists on sending the best response an agent (taking the role of advisor) can give to the problem being solved considering the state of the problem as seen by another agent (the advisee). The advisee uses this information as input to a supervision mechanism. The way the advisee uses information depends on the expected quality of the advice given by the advisor and a measure of trust kept by the agent based on the result of previous advice given by the same advisor. Several other mechanisms, such as providing unsolicited advice and combining advice, are under study.

• a simpler traffic light control system including adaptation has also been developed and is reported in [Dia03].

CURRENT AND FUTURE WORK:

• We are now evaluating this method on the simulated traffic control problem. The simulation is similar to the one used by the Gerhard-Mercator-Universität Department of Physics of the University of Duisburg (Germany) to study traffic phenomena. This simulation has been built using Java2 and C++, running on Linux.
• Our approach and results have been published and extensively discussed in the "Adaptive Agents and Multi-Agent Systems" AAMAS-3 workshop collocated with AISB'03 conference [Nun03e], and the interest generated by this approach lead to the publication of several other works including an article in a journal [Nun03a] and a book chapter [Oli03a]. The paper [Nun03c] got a nomination for the best paper at CIA'03.
• The tests are now being done about the interaction between agents that use Reinforcement Learning and Evolutionary Algorithms, although the approach is, in many respects, generalizable to other learning algorithms. Simulation uses real traffic data from a portuguese town. We want to be sure about "poor" agents performance improvement at the level of the best ones as well as whether agents interactions become stable or not.
• Current and future work will focus on refining the advisor selection process, based on the concepts of mutual trust and self-confidence, concepts that dynamicaly evolve with the agents experience. Identifying teams of agents and influencing team members on the choice of a specific advisor will also be tried out through advice and information exchange. It may also be interesting to see if team supervisors will emerge from the interactive learning process. Other issues include combination of advice from several different sources and dealing with unsolicited advice and influence. Extending this approach to other different learning algorithms and testing its effectiveness in problems of traffic-control simulation will be done. Other problems, such as communication networks and band-width management are under study as possible applications.

Numerical Reasoning in Inductive Logic programming (ILP)

Research goals: To enhance an existent Inductive Logic Programming learning system with numerical capabilities.

Inductive Logic Programming (ILP) has been very successful in applications involving classification problems. ILP algorithms may easily take advantage of expert provided information relevant for the induction task. Theoretically there are no constraints on the nature of such expert information. Having that in mind we consider the possibility of collecting and making available to an ILP system a set of libraries of numerical and statistical methods. These library may be used by an ILP system in any application requiring numerical reasoning. With a proper capability to handle numerical applications ILP may widen the range of successful applications where it can by used. FUTURE WORK:

• The goals of this work are two folds: i) development of a set of libraries incorporating numerical and statistical methods relevant for applications requiring numerical reasoning; ii) adapting and extending existing ILP systems to improve their performances in such applications.
• The first line of research includes the development of general purpose libraries for numerical reasoning and also more specific ones adequate for specific application with great practical interest like Temporal series. The second line of research includes substantial improvements in current ILP systems. It will be necessary to include in ILP systems statistical methods and changes in the search methods used.
• We will identify the set of interesting applications where our developments will be evaluated.

Emotion-based Agents

Research goals: The second question to be answered is: Will it be possible to escape from usual utility-based decision functions in which decision-making for autonomous agents is concerned?

We are pursuing research efforts towards the understanding how emotion-like concepts can endow agents' architecture with more sophisticated, although still reliable, decision-making capabilities. New agents' architectures have been proposed and experiments are being done in simple simulation scenarios.

Although the study of emotion in the realm of Artificial Intelligence is not totally new and has been addressed by Simon, Minsky, Sloman and Croucher among others, recently much more attention has been devoted to this subject by several researchers like R. Picard. This renewed effort is being motivated by trends in neuroscience (see A. Damásio's recent work), that are helping to clarify and to establish new connections between high level cognitive processes, such as memory and reasoning, and emotional processes. These recent studies point out the fundamental role of emotion in intelligent behaviour and decision-making. From our perspective, as engineers and computer scientists, we are mostly interested in studying the functional aspects of emotional processes. Particularly, we aim at understanding how emotional mechanisms can improve cognitive abilities, such as planning, learning and decision-making, for hardware and software Agents.

RECENT WORK (2003):

• After having proposed what we believe to be an innovative emotion-based agent architecture we came to the conclusion that, in order to make experiments that could be evaluated, we needed to develop a software platform enabling an apropriate simulation. The main requirements for this needed simulation environment were: (1) high level of complexity for the participating Agents; (2) real-time, highly dynamic, computational demands; (3) multiple concerns at stake;(4) overall goals asking for a Multi-Agent System approach; (5) closeness to a real-world problem.
• Most of our recent work was then focused on the development of the Pyrosim simulation platform, with the aim of creating a suitable simulation environment for the development of Emotional Agents. The Pyrosim platform simulates a forest environment where a team of Agents is placed to fight an ongoing fire. Agents representing fire fighters deal with all kind of realistic constraints (water availability, their own energy and acceleration, vegetation type and density, terrain slope, wind...). The platform besides providing all the simulation environment also alows agent communication. The core of the Pyrosim Platform includes two basic elements:

  1. the Pyrosim Server application, responsible for running all the simulation logic. It contains the environment's model and updates the state of every entity in each simulation cycle.
  2. the Agent Skeleton layer that provides a convenient interface for creating customized Agents. It manages all low level communication between a customized Agent and the Pyrosim Server.

• There is also another application that is very useful for monitoring the state of the simulation: the Pyroviz. The Pyroviz allows the developer to visualize almost in real-time what is happening during the entire simulation through a 3D rendered scene. The Pyrosim platform does not provide any Emotional Mechanisms.
• We have been proposing an agent architecture that includes several emotional-like mechanisms, namely: emotional evaluation functions, emotion-biased processing, emotional tagging and mood congruent memory. These emotional-like mechanisms are intended to provide agents with increased performance and adaptability in real-time environments. Previously, we have divided emotional phenomena in 3 different categories: specific emotions, moods and emotional dispositions. We are now concerned with the dynamics of emotional mechanisms. Emotional Accumulators (EA) couple with Emotional Evaluation Functions (EEF) to build another element of our agent's architecture, the Basic Emotional Structure similar to well known emotions ("Fear", "Anxiety", "Self-confidence",...). We want to enable dynamic variations of those parameters that influence the agents' planning capabilities to achieve specific goals. There is a very strong link between emotion and memory, a link that has several functionalities associated. One of them is emotional tagging. Instead of being simply stored in memory, information is first marked with an emotional tag, relating that information with the current emotional state of the agent. The tagging procedure is itself important because it allows selecting which information should be stored either in long-term memory or in short-term memory or simply discarded.
• A revised implementation of Emotion-base Architecture was started, following a more refined version of the model previously developed. Details about the current model are given in [Oli03b].

• Future work will be focused in developing a complete implementation of the Emotion-based Architecture, with the Pyrosim scenario in mind. Key concepts on Emotion-based Architectures are:

• (i) Emotional Evaluation Functions;
• (ii) Emotional State;
• (iii) Emotional Accumulators;
• (iv) Basic Emotional Structures;

• The next step will then be experimenting and testing specific Emotional Mechanisms using the Pyrosim simulator. We will be trying to test how Emotional Mechanisms such as "Fear", "Anxiety" and "Self-Confidence" may be used as:

• Information for internal processes (Emotion as Information)
• Changing internal processing parameters (Emotion as a Process Control Mechanism)
• Balance computational effort and manage resource allocation (Emotion and Resource Allocation Mechanisms)
• Generate global processing modes (Emotion and Information Processing Strategies)

D.1.3.3 Multi-Agent Coordination and Cooperative Robotics

Research direction: Coordinating teams of autonomous (or semi-autonomous) agents that perform in rich, dynamic, both competitive and adversarial environments is a major aim of this work line. For this objective, we are exploring several research directions that can be seen as complementary:

• (i) Developing new coordination protocols that enable teams of agents to perform complex tasks in open multi-agent environments
• (ii) Organising and making available knowledge, languages and protocols for enabling agents' teamwork;
• (iii) Defining methodologies for analyzing team behavior and to use this analysis for coaching teams of autonomous agents;
• (iv) Design and implement an agent-based common framework suitable for controlling teams of cooperative robots (either physical or simulated) playing in RoboSoccer-like competitions or performing other team complex tasks in dynamic domains;
• (v) Development of realistic multi-agent simulators enabling the analysis of team coordination methodologies for performing complex tasks in dynamic environments as it is the case with a Coastal Ecosystems Simulator
• The research conducted along this line, may be also partially applied to several other very useful social areas, like mine clearance, land exploration, control of hospital robots, public transport coordination, satellite control, battlefield scenarios, cleanup of radioactive and toxic contamination and other problems that imply team coordination in dynamic environments

Conflict Resolution in Decision Support Systems

• Previously concluded work in this research direction, includes DiPLoMAT prototype (Distributed Project Location Multi-Agent based Testbed). DiPLoMAT applies most of the theoretical work on conflict resolution done before, to the Project Location assessment domain. Evaluation Agencies have to identify the project land requirements (area, natural resources, transports, price, etc.), and finding an adequate location for that specific project. The achieved result is an Intelligent Decision Support System to assist the project team in finding adequate projects' locations which comply with the applicable legislation and satisfies the project requirements.

Multi-Agent Teams' Coordination  

RECENT WORK (2003):

• Simulated Soccer Team: Our Simulated soccer team - FC Portugal (developed in collaboration with IEETA/Univ.Aveiro), ranked 5th in RoboCup 2003 competition held in Padova, Italy;
• Coach Agent: Our Coach agent - FCPortugal Coach - ranked 2nd in the 2003 coach competition in the RoboCup competition that took place in Padova;
• Other RoboCup teams: We have developed some simple methodologies for coordinating teams of physical robots and applied them in the context of 5dpo RoboSoccer small-size and middle-size league teams (developed by ISR-Porto).
• Development of a Sony legged RoboSoccer team in the context of Portus and LEMAS projects (together with ISR-Porto). Our Legged league team FC Portus, ranked 5th in RoboCup 2003 held in Padova;
• Coordination of Teams of Homogeneous/Heterogeneous Agents in Adversarial Environments. We proposed several coordination methodologies for dynamic spatial domains, namely: Strategic coordination, Situation Based Strategic Positioning (SBSP), Dynamic Positioning and Role Exchange (DPRE).
• Coordination using High-Level Communication. We have proposed Intelligent Communication policies based on appropriate (multi-level) World State knowledge(ADVCOM) together with a way of balancing strategic and active communication.
• Coordination without Communication. We have proposed some opponent modeling and teammate modeling techniques enabling coordination without communication. We also have developed a more adequate perception mechanism called Strategic Looking Mechanism (SLM) for coordination without communication
• Partial Hierarchical Coordination. We have developed techniques enabling a team of autonomous agents to integrate coach advice in their reasoning processes, together with their local reasoning process.
• A Visual Debugger for analysing intelligent agent-based players was improved and released as well as COACH UNILANG, a standard Language for Coaching a (Robo) Soccer Team.
• A new graphical tool called Team Designer enabling easy definition of a complete soccer strategy by a human designer was developed. This tool has a simple automated coach, a game analyzer and a graphical visualizer merged into a complete coaching environment.
• Multi-agent realistic camera control was introduced in our Graphical 3D Visualizer for simulated soccer games.

CURRENT AND FUTURE WORK:

• Definition of new coordination protocols for our simulated soccer team (in the context of FC Portugal - New Coordination Methodologies Applied to the Simulation League Project - FCT/POSI/ROBO/43910/2002);
• Definition of new methodologies for intelligent perception, teammate and opponent modeling and multi-agent communication (in the context of FC Portugal project and LEMAS - Learning in Multi-Agent Systems in the RoboCup Sony Legged League project - FCT/POSI/ROBO /43926/2002);
• Enhancing the definition of a COACH UNILANG as a general language to enable a special agent ("coach") to supervise a team of co-operative robots.
• Research on opponent modeling and automated coaching and enhancement of our automated coach with intelligent resource allocation and learning abilities for tactical selection;
• Developing a Sony legged RoboSoccer team in the context of Portus and LEMAS projects (together with ISR-Porto).
• Development of multi-agent learning methodologies for selecting the best coordination methodology for each situation;
• Increasing the co-operation with ISR-Porto (in the context of 5DPO, Portus and LEMAS projects), IEETA-Aveiro (in the context of FC Portugal project) and CEMAS-UFP for the developement of a more realistic multi-agent ecological simulator.

D.1.3.4 Other Agent-based systems applications

People involved: Alexessander Alves, João Luís Pinto,Hugo Proença, Guilherme Pereira, Rui Camacho, Eugénio Oliveira

Research direction: To apply agent architectures, negotiation protocols and learning algorithms to specific application domains.

Communications network management

Research goal: To apply learning algorithms to improve resource allocation in multi-class packet switched networks. The system will be applicable to multi class settings where decision-making problems are exceptionally complex.

RECENT WORK (2003):

• We have tried several different methods for Time Series Analysis to deal with all the past information on traffic control in multi-class packet switched networks. The objective was to conceive a learning system that adapts to changing traffic conditions and predict future traffic conditions based on past traffic patterns. Both Deterministic Linear and Non-Linear methods as well as Stochastic methods have been tried out. Most recent work has been done in order to accessadequacy of Inductive Logic programming (ILP) for Time Series Analysis automation in the context of traffic engineering of data Communications Networks [Alv03b]. Our proposals included improving ILP systems with numerical reasoning capabilities as well as to provide those ILP systems with apropriate background knowledge for Time Series Analysis. We may say that results were improved when compared with statistical Time Series methods [Alv03a]. IndLog, an ILP system previously developed by R.Camacho, and YAP prolog language were the basic software tools in which our modifications and extensions were included.

• We are now applying our Inductive Logic Programming algorithm [Alv03a] to time series data. The models generated by the ILP algorithm were expectd to give accurate predictions of the observed data over the "lead-time" period and they also should contribute to the human understanding of the network behaviour. Many experiments have to be done. A complete multi-agent system managing the allocation of all available resources in multi-class packet switch networks and using the induced prediction capabilities will be specified and, hopefully, implemented.

Electrical Energy e-Market

Research goals: Current European efforts for the establishment of both de-regulated Electrical Energy Markets and Electronic Commerce platforms can be brought together through appropriate multi-agent platforms for autonomous agents trading interaction.

RECENT WORK (2003):

• In the specification of the multi-agent system encompassing the needed functionalities for the Electrical Energy e-market, we have untill now emphasising security procedures, accountability of the communications, good performance and software portability. Also, integration with legacy systems has been privileged.
• The Agents' architecture has been proposed and all the interaction protocols specified. Giving the choices of using XML and FIPA-ACL, it was settled that message transfer should be made using HTTP POST requests (both from each agent to the facilitators and from this one to specific advisor agent). Also Client requests are made using CET requests. All interactions use MIME type text/XML. Market operator acts as a message router.
• There are two different "operators" in the system: Market operator who manages the trading interaction between the agents in the market, and the System operator managing the market interaction with the power system, checking for the feasibility of the market operation. The kind of interaction we have implemented simulates a spot market where bids are on power, marginal costs and hour of the day.
• In our Electricity E-Market, agents authenticate through digital certificates, while messages between the market operator and the market agents are digitally signed. We have selected the TLS/SSl protocol and, as for the message digital signatures is concerned, the open standards are being used. They rely on classical MAC and cryptography algorithms used in SSL. Market operator is seen as a trusted third partner, responsible for registration, auctions and matching bids and offers.

• We are now in the process of implementing a prototype for the Electrical Energy e-Market agent-based system. A first scenario will be specified to check secure and reliable agent interaction. A secure Multi-Agent System framework making available negotiation algorithms tuned for this specific domain is to be achieved. We foresee the implementation of a prototype to showcase all the technological options. Open source tools and libraries are being used.

Controlling a Metropolitan Railway System through Agents

Research & Development goals: To specify and implement a generic multi-agent system suitable for automatic control of a subset of a railway system.

RECENT WORK (2003):

• During the year 2003 we have specified and implemented MARCS (Railway control Multi-Agent System) that proves, in a simulated environment,that by using communication it becomes possible to take correct control decisions and potential conflicts may be avoided. A Master thesis report was produced [Pro03]. MARCS makes agents representing trains, train stations and sub-network supervisors, to cooperate in order to preview whether or not conflicts can occur. A conflict arises when a resource, here a joint in a railway line or a station train-stop, is simultaneously required by more than one train.
• MARCS is a decentralized control system that better deals with local exceptions and also learns to adapt to future potential conflicts [Pro03a]. Agents (trains, train stations, sub-network supervisors) exchange either plans or other specific information to help in their own reasoning to coordinate their speed and avoid potential conflicts. The system becomes fault-tolerant, once whenever a sub-network supervisor is down, it immediatly reconfigures itself to re-distribute the responsibility by other supervisor agents.
• Each potential confict that has been solved by the system is then recorded. The all set of previous examples (avoided conflicts) is dealt with by means of an unsupervised learning algorithm - APRIORI (Agrawal, Strikant)to induce rules of behaviour for future application in similar situations.

CURRENT AND FUTURE WORK (2004):

• We intend to finalize this work by evaluating the results (have been potential conflicts correctly antecipated?) after a number of different experiments in the simulated network provided by MARCS. Also an analysis on the possible generalisation to different, although closer application domains, will be provided.

Agent-based Travel Services platform

Research & Development goals: To specify and implement a generic multi-agent system platform suitable for supporting automatic Travel Services Agency. This research has been motivated by the "Personal Travel Assistant" application from British Telecom and the Trading Agent Competition.

RECENT WORK (2003):

• We have specified and implemented a platform combining the distributed nature of a multi-agent system with the security and encryption facilities provided by web servers, separating data from knowledge in a concurrent common area. We have also endowed the agents with learning capabilities directed to two different aspects: Reinforcement Learning for bidding policies and a kind of Concept Learning technique for user profiling. The functionality of all the proposed multi-agent system features is now being illustrated through an agent-based Travel Services application.
• The system allows flight and hotel booking from local or remote customers. There are three main types of agents: customer, company and hotel agents who represent the respective entities. There is also a facilitator, a press agent and a softbot. The facilitator is responsible for storing travel packages bundled by company and hotel agents. The press agent is an agent who will publish customer complaints as well as good impressions. It serves as a mean to establish company and hotel agent reputation. The softbot is capable of extracting information from real world web pages updating company and hotel agent information systems.

• This work will be complete after experimentation in scenarios where gains from the agents learning features, concerning both user profiling and negotiation adaptation, can be highlited.

D.1.3 NIAD&R: Plan for 2004