NIAD&R Distributed Artificial Intelligence & Robotics Group • NIAD&R is a member of LIACC Artificial Intelligence and Computer Science Laboratory University of Porto Overview of NIAD&R Coordinator: Eugénio Oliveira http://www.fe.up.pt/~eol/MEMBERS/eco_html LIACC - March 2004 1 NIAD&R Distributed Artificial Intelligence & Robotics Group OUTLINE LIACC overview NIAD&R main Goals NIAD&R through Numbers Main research lines 1 Models for Agents’ interaction 2 Advanced features for Autonomous Agents 3 Coordination of Agent-based teamwork 4 Agent-based Applications Projects Conclusions Future work LIACC - March 2004 2 NIAD&R Distributed Artificial Intelligence & Robotics Group ~50% NCC F. Sciences University of Porto Advisory Committee ~23% Coordination Board NIAD&R F. Engineering general coordinator ~27% NIAAD F. Economics LIACC - March 2004 Scientific Board 3 NIAD&R Distributed Artificial Intelligence & Robotics Group Existe desde 1988. No fim de 2003 incluía: 34 pessoas com doutoramento, 40 outros membros (estudantes de doutoramento, bolseiros, etc.) LIACC - March 2004 4 NIAD&R Distributed Artificial Intelligence & Robotics Group Nos últimos 5 anos (‘99 – ‘03) foram publicados: • 30 artigos publicados em Revistas Internacionais submetidos a avaliação independente (13 em 2003); • 105 artigos em livros publicados pelas editoras reconhecidas (ex. Springer), submetidos a avaliação independente, e com entrada no “Science citation index” • • • • 1 livro (editado); 16 capitulos de livros; 133 outros artigos in actas de Congressos / Workshops 9 actas de congressos ou Workshops editadas; LIACC - March 2004 5 NIAD&R Distributed Artificial Intelligence & Robotics Group (’99 - ‘03) foram orientadas e concluídas: • 18 Teses de Doutoramento, • 34 Teses de Mestrado. •Em 2004 estão a ser orientadas: • 21 Teses de Doutoramento, • 37 Teses de Mestrado. LIACC - March 2004 6 NIAD&R Distributed Artificial Intelligence & Robotics Group Nos últimos 5 anos (’99 - ‘03) foram organizados: • 19 congressos / workshops Em 2003 os membros do LIACC participaram em: • 22 comissões de programa • 6 editorial boards de revistas LIACC - March 2004 7 NIAD&R Distributed Artificial Intelligence & Robotics Group NCC: - Declarative programming Logic Programming Systems Parallel Execution of Logic Programs Constraint Programming - Parallel and Distributed Systems Concurrency, Distribution and Mobility OO Languages for Distributed Environments Parallel Programming Environments - Logic, Language and Computation Formal Systems; Logic and Grammars Complexity - Automatic Evaluation of Students Exercises - Geo-Referenced Data Processing LIACC - March 2004 8 NIAD&R Distributed Artificial Intelligence & Robotics Group NIAAD: Data Mining and Decision Support Collaborative methodologies for DM & DS. Recommendation tools for the selection of DA methods Modeling Dynamic Systems Modelling complex dynamic systems, Advanced Techniques in Data Mining and DA Modelling higher order concepts using ILP Statistical methods for classification. App. to the recognition of multi-spectral images and bioinformatics. Data Mining for Text and Web LIACC - March 2004 9 NIAD&R Distributed Artificial Intelligence & Robotics Group NIAD&R Main Objectives: To develop: Models for inter-operability in Agent-based Systems Agent-based Software for practical Applications in DDD To help young researchers in preparing their thesis LIACC - March 2004 10 NIAD&R Distributed Artificial Intelligence & Robotics Group Main Concept: AGENTS : Computational entities including the following features: Distributed Reactive and Communicative Pro-Active Autonomous Other capabilities: Mentalistic capabilities: Beliefs, Desires, Intentions, Emotions Multi-Agent Systems LIACC - March 2004 11 NIAD&R Distributed Artificial Intelligence & Robotics Group What is the rationale behind our research? Agents Interaction Negotiation protocols Coordination of teams of Agents Conflict resolution Agent Capabilities Agent Applications LIACC - March 2004 Learning and Adaptation Emotion-based agent architecture Electronic Institutions for VO Multiple-Experts DSS Simulation 12 NIAD&R Distributed Artificial Intelligence & Robotics Group Nr. of Researchers Senior Res. PhDs Res.Ass. Ext.Coll. TOTAL Tech. Adm. LIACC - March 2004 1999 2000 2001 2002 2003 1 1 5 2 9 _ 1 2 6 2 12 _ 1 4 13 3 21 _ ½ ½ 1 2+1-1 8-1 2 11 _ 1 4 13 3 20 _ ½ (shared) ½ (shared) ½ (shared) 13 NIAD&R Distributed Artificial Intelligence & Robotics Group Scientific production 1999 2000 2001 2002 2003 Chap. in Books Journ + Series Proceed +TRep Total Publications 1 +4 4+6 9 (+6) 4 1+3 2*+5 3+9 13+8 11 (+9) 20(+8) 1 1+9 9+4 20(+4) PhD Thesis approved 1 MSc.Thesis approved 4 2 1 1(+1) 2 1 1 1 2 Total (Theses appr.) 5 3 4 2 3 (without Tech. Reports) Theses in prep(PhD+M) 1 1+3 15+1 20 (+1) 4+4 3+8 5+6 * Interview to TRN- Technical Research News Magazine LIACC - March 2004 14 NIAD&R Distributed Artificial Intelligence & Robotics Group Scientific production 1999 2000 2001 2002 2003 Prototypes (rev) 2(+3) 2+(2) 3+(1) 4+(2) 2+(4) Conf. Org.+PCs +4 +5 3+5 1+10 1+11 Ed.Boards 1 1 2 2 2 2 3+4 1+4 7 2 3 2 1+* P.g.Courses+Inv. Sem. Awards (prizes) - * Plus one nomination for the best paper at CIA’03 LIACC - March 2004 15 NIAD&R Distributed Artificial Intelligence & Robotics Group Main Research lines: 1. Flexible and trustful tools and platforms for agents interaction: Electronic Institutions for B2B LIACC - March 2004 Automatic negotiation Distributed Belief Revision 16 NIAD&R Distributed Artificial Intelligence & Robotics Group 1. Flexible and trustful environments for agents interaction: Norms & Rules V.O. formation V.O. operation Q-Negotiation Monitoring links to other Institutions V.O dissolution Ontology Services EAgt financial MAgt legal LIACC - March 2004 EAgt EAgt EAgt Electronic Contract Ana Paula Rocha+ Henrique + Andreia+ ECO 17 NIAD&R Distributed Artificial Intelligence & Robotics Group Negotiation and Electronic Institutions Q-negotiation protocol: Multi-Attribute bid Evaluation Qualitative feedback Adaptive bid formation Distributed dependencies resolution FOREV implementation (V.E. formation platform) Electronic Contracts (in progress) Ontology Services (in progress) LIACC - March 2004 Protocols for both inter and intra coalition negotiation for Distributed Resources management MACIV (MAS for Civil Construction) ECO + JMFonseca-UNL 18 NIAD&R Distributed Artificial Intelligence & Robotics Group Agent-based System for EC : MAS architecture suitable for B2C interaction Agent Tactics and Strategies for Negotiation Adaptive to the market dynamics (Q learning) SMACE prototype available through the WEB Time-dependent tactics Utility Utility Utility 0 0 0 Time LIACC - March 2004 1 1 1 Greedy Linear Anxious Time Time Henrique Cardoso + ECO 19 NIAD&R Distributed Artificial Intelligence & Robotics Group I can do 70 for 40 I can do 30 for 20 I need 100 I can do 50 for 30 I can do 90 for 50 D Need 100 30 for 20 Task announcing 70 for 40 50 for 30 Coalition formation Edification Machinery Supervisors Stocks manager Workers Storehouse Four useful coalitions: - C & D (cost=70) - B & C (cost=60) - A & C (cost=90) - A & D (cost=80) Coalition team & best offer announcing to coordinators Leading Quit Facilitator Stocks manager One software agent for each physical resource ... 55 New best offer announcing to coordinators Leading 45 Leading Quit Chief Intra-coalition negotiation 50 The winner !!! Facilitator 80 Task announcement & 100negotiation100 Garage 75% 100 OK A B 40 20 Initial costs Facilitator C 30 D A B 10 Coalition configuration Machinery C 100 64% Personnel A B C 100 62.5% C D A B A OK but D refuses 75 100 Facilitator D 7 A B D Coordinator proposes new percentage to A and D 25C A and D OK but B and C refuses 100 OK Workers OK A B 18 D Coordinator receives new lower value and divulges the lowering percentage C D Coordinator proposes new percentage to A 75 OK A B C D A B A accepts C D Coordinator sends new proposal Intra-coalition negotiation Planning Experts Facilitator LIACC - March 2004 Edification Garage Storehouse Planning Personnel Interconnecting network A 1 Coordinator 1 B 2 C Coordinator 2 The self-depreciation problem JMFonseca + ECO 20 NIAD&R Distributed Artificial Intelligence & Robotics Group Multi-agent systems dealing with Conflict resolution Argumentation-based Belief Revision in a Distributed MAS involving a GIS Application domain: DSS for Land use assessment Prototype: DiPLoMAT System LIACC - March 2004 available M.Benedita Malheiro + ECO 21 NIAD&R Distributed Artificial Intelligence & Robotics Group Cooperation Layer Self Model Cooperation Layer Expert System Acquaintance Model Communication Module Cooperation Module Assumption based Belief Revision System Problem Solver ATMS Knowledge Base Facts Assumptions Justifications Consumers Beliefs Scheduler LIACC - March 2004 Inferred Nodes 22 6. DOMÍNIO DE APLICAÇÃO NIAD&R Distributed Artificial Intelligence & Robotics Group LIACC - March 2004 23 NIAD&R Distributed Artificial Intelligence & Robotics Group Main Research lines: 2. Advanced features for Autonomous Agents: LIACC - March 2004 Agents and MAS Learning capabilities Learning marketing strategies “Emotion-based" agents’ architectures 24 NIAD&R Distributed Artificial Intelligence & Robotics Group 2. Advanced features for Autonomous Agents: Agents and MAS Learning capabilities “How can Heterogeneous Agents interactively learn and “influence” each other in their learning process?” Non-deterministic, partially observable, Nonsupervised environment- Traffic Lights Control LIACC - March 2004 Luis Nunes + ECO 25 NIAD&R Distributed Artificial Intelligence & Robotics Group • Traffic Light Control • Simulation based on real data • Simplified car movement • 1, 2 and 4 crossings scenarios x 3 teams (GA, QL, Heuristic) LIACC - March 2004 Luis Nunes + ECO 26 NIAD&R Distributed Artificial Intelligence & Robotics Group Learning for Multi-Agent Systems: Exchanging Advice Use communication to improve learning performance • Environments: Multiple agents dealing with similar problems Agents use different learning techniques • Expected features: Improved resistance to local minima No pre-selection of the “best algorithm” Group performs better than the best of its individuals LIACC - March 2004 Luis Nunes + ECO 27 NIAD&R Distributed Artificial Intelligence & Robotics Group 3.To whom ? Environment T7.A,Ar * PAr(S) advice > TA,O * PO(S) Integrate 1. Observed State Advisee 4. 8. State (observed theact Advisee) Choose actionby and 2. Should I request advice ? ScA * PA < PO or ConfusedAbout(State) Advisor LIACC - March 2004 Self-confidence Performance State Trust Actors: 6. Advised action (a) 5. Best guess is (a) Concepts: Advisee Advisor Other agents 28 NIAD&R Distributed Artificial Intelligence & Robotics Group “Emotion-based“ agent architectures: based on neuro-science (cognition-emotion relationship) “How to escape from traditional utility-based functions”? Agents new features: valence-based memory, <V,I,E,G> associating Valence with Goals and both internal and external sources Relationships between “emotion-like parameters” Luis Sarmento + Daniel Moura+ECO LIACC - March 2004 29 NIAD&R Distributed Artificial Intelligence & Robotics Group • Emotional Elicitation – Process of evaluating the chances (V) of achieving a given goal (G) upon the state of the environment (<E>) and the agent internal state (<I>). V = EEFG(<E>,<I>) • Emotion Accumulators – Enable to model emotions behavior through time (t); – Consumes a percentage (PInput) of an EEFG; – Value decays according to a decay constant (Td). EAG(t,PInput ,Td) • Basic Emotional Mechanism EEFG(<E>,<I>) Luis Sarmento + Daniel Moura+ECO LIACC - March 2004 EAG(t,PInput ,Td) 30 NIAD&R Distributed Artificial Intelligence & Robotics Group Main Research lines: 3. Coordination of Agent-based teams LIACC - March 2004 Coordination policies in adversarial environments 31 NIAD&R Distributed Artificial Intelligence & Robotics Group 3. Coordination of Agent-based teams agent-based common framework suitable for controlling teams of cooperative robots (either physical or simulated) Techniques: distinction between active and strategic situations Agents Coordination mechanisms: Situation-basedSP, DPRExchange, ADVCOM, SLM, MM. New team strategies (tactics, formations, player types…) COACH UNILANG: general language to enable a special agent ("coach") to supervise a team of co-operative robots. LIACC - March 2004 Luis Paulo Reis 32 NIAD&R Distributed Artificial Intelligence & Robotics Group Flexible Strategies LIACC - March 2004 33 NIAD&R Distributed Artificial Intelligence & Robotics Group Constraint Satisfaction in a distributed environment: UNIPS - University Planning and Scheduling to reach mutual agreement in distributed multiagent system applications UniLang: language for representing timetabling problems LIACC - March 2004 Luis Paulo Reis 34 NIAD&R Distributed Artificial Intelligence & Robotics Group Main Research lines: 4. Application oriented: Proof of “intelligent agent” concept in specific application domains: Elec. Market, Mob.Comm.Networks, Brokering, 3D Visualisation, Traffic Control management… LIACC - March 2004 35 NIAD&R Distributed Artificial Intelligence & Robotics Group 4. Application oriented work: Intelligent Brokering for the Insurance domain (BeeGent) – LNogueira ILP for Time Series Analysis. Optimal traffic control in multi-class packet switched networks –Alex Alves MAS platform for Electrical Energy e-Market - JLPinto Agent-based framework making available security mechanisms and negotiation algorithms tuned for this EEmarket LIACC - March 2004 36 NIAD&R Distributed Artificial Intelligence & Robotics Group 2-BIAS (Brokerage in Insurance – an Agent-based System) Interaction on user’s behalf Customer description and needs CA Product proposals Broker IA Customers communities IA Stereotypes ... Negotiation Personalised offers LIACC - March 2004 Qualitative feedback IA Ontology Luis Nogueira+ECO 37 NIAD&R Distributed Artificial Intelligence & Robotics Group Traffic Engineering of Data CommunicationsNetwork VBR Traffic Time Series Forecasting observed 90 85 80 75 Load –To assess the adequacy of ILP for Time Series Analysis automation predicted 70 65 60 55 50 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 Tim e LIACC - March 2004 Alex Alves+RC+ECO 38 96 NIAD&R Distributed Artificial Intelligence & Robotics Group Multi-agent Platform for Electricity E-Market • Agents interaction using XML and HTTPS (or SOAP?) • Intra-platform communications with XML-RPC • Market Operator acts as a message router; Auctions. • Agents authenticate through digital certificates • Messages between the Market Operator and the Market Agents are digitally signed LIACC - March 2004 Certificatio n Authority Database Auctioneer Market Operator System Operator System Service Provider Platform Generator Agent Consum er Agent Service Provider Agent João Luis Pinto Service Consumer Agent 39 NIAD&R Distributed Artificial Intelligence & Robotics Group MAS for 3D visualisation of RoboCup games For the RoboCup Simulation League Agent-based control Cameras deciding on the best perspective on the situation + Director agent Sérgio Louro+LPR+ECO LIACC - March 2004 40 NIAD&R Distributed Artificial Intelligence & Robotics Group Agent-based Robotics control: Hybrid layered architecture: Reactive Agents for immediate action Deliberative agents responsible for advising about the plan to be execute Learning basic competencies : Simple neural networks, Fuzzy rules 1 PhD Thesis submitted and several papers produced LIACC - March 2004 41 NIAD&R Distributed Artificial Intelligence & Robotics Group Agent-based Robotics control: LIACC - March 2004 42 NIAD&R Distributed Artificial Intelligence & Robotics Group Projects (1999-2001) AVOEC: Agent-based platform for VE formation and B2B EC funding:FCT+FEDER finished 2001 MACIV:Multi-Agent System for Distributed resource management funding:FCT+FEDER finished 2000 AgentLink I (98-00) AMEC SIG finished 2000 AgentLink II (00-02) AMEC +ALAD SIGs funding: European Union FINESSE: Formalisation of Institutions and Norms for Electronic Social Structures for Exchange submitted Inter-Network (EUNET, ILPNET, AgentLinkII) SIG on "Agents that Learn, Adapt and Discover funding:E.U. LIACC - March 2004 43 NIAD&R Distributed Artificial Intelligence & Robotics Group Projects (2002-2004) LEMAS: Learning in MASystems in the RoboCup SLLeague Funding: FCT+FEDER(03-04) FCPortugal: New Coordination Methodologies applied to the Simulation League Funding: FCT+FEDER(03-04) AgentLink III AMEC +ALAD SIGs Funding: European U. (03-) PORTUS: A common framework for cooperation in Mobile Robotics Funding: FCT+FEDER(02-05) OPEN: Open Platform for Enterprise Network – submitted to E.U. LIACC - March 2004 44 NIAD&R Distributed Artificial Intelligence & Robotics Group International links Member of the Editorial Board of the AAMAS Journal ed. Kluwer AP, (EO) Member of the European Board of IOS Press and Omsha Ltd “Frontiers in AI and Applications” Series for European dissertations (EO) Member of Technical Committee of the RoboCup Simulation League (LPR) Exchange of students+researchers under Socrates Program (U.Trier/G, Imperial College/UK, City College/UK, ENM SaintEtienne/Fr) Research Evaluation at Univ. J.Fourrier-Grenoble/Fr “Coordination and Cooperation” in MAS Robo Cup-SIG LIACC - March 2004 45 NIAD&R Distributed Artificial Intelligence & Robotics Group International links DFKI-Germany (KlausFisher, Mathias Klush) Imperial College/Univ.London (A.Mamdani) Czech Technical University (V.Marik, O.Stepankova) U.Southampton (N.Jennings, M.Luck) Lab. Leibnitz- IMAG (Dr. Y.Demazeau) Univ. Utrecht (F.Dignum) Université de Technologie de Compiègne (J.P.Barthés) École National des Mines Saint-Etienne (O.Boissier) Institut Inteligencia Artificial, Barcelona (C.Sierra) QMWC/U.London (N.Jennings) LIACC - March 2004 46 NIAD&R Distributed Artificial Intelligence & Robotics Group International links Univ. São Paulo (J.Sichman) Pontifícia Univ. Católica Paraná (M.Schmeil) Univ. Bath (J.Padget) Univ. Federal Rio Grande do Sul (A.Bazzan) Univ. of Trier/Germany (N.Kuhn) Xerox Research Centre in Europe, (J.M.Andreolli) Achmea, Netherlands (V.Dignum) LIACC - March 2004 47 NIAD&R Distributed Artificial Intelligence & Robotics Group National links Faculdade de Ciências Univ. Lisboa (Prof.H.Coelho) Instituto Superior Técnico, U.T.Lisboa (Drª. A.Paiva) FCT- Univ. Nova de Lisboa (Prof. A.Garção) IEETA- Univ. Aveiro (Dr. L.Seabra Lopes, Dr. N.Lau) INESCPorto (Prof.M.Matos) ISR Porto (Dr.A.PMoreira, Dr. P.Costa) CEMAS-C.Modelação e Análise Sistemas Ambientais (Prof. P.Duarte) I2S- Integrated Systems Software Company(A.Lhamas) Guião (J.A.Alves) Mota&Companhia (Civil Construction) Univ. Beira Interior (DrªP.Prata) LIACC - March 2004 48 NIAD&R Distributed Artificial Intelligence & Robotics Group Weak points: Difficult to keep fruitful links with industry Not enough publications in Journals Robotics not enough attractive for software people “Strong” points: EI and Agent-based Negotiation for EC and VO Sophisticated prototypes have been released Good results in competitions by using agent-based team coordination New research directions like “emotion-based” agents and MAS Learning LIACC - March 2004 49 NIAD&R Distributed Artificial Intelligence & Robotics Group Future Directions: More efforts on: Multi-agent Learning Emotion-like Agents Architectures General framework for Electronic Institutions including Ontology related Services and Electronic Contracts More efforts on the Applications: Adaptive Negotiation Electricity e-Market Simulation Tool for fire-combat training Application of EI to a real-life domain To keep a fair balance between Research and Applications LIACC - March 2004 50