Service agents n n n Publish white page

Service agents n n n Publish “white page” services description content and register the services at a “yellow page” site Understand ontology and answer queries Link with the semantic web server and push information to other agents May 6, 2004 1

Features n n n Distributed: no centralized agent who has to search all web pages and understand every ontology The best agent to ask questions can be easily located: a good amendment to the web services discovery and the agent services searching The non-semantic web site joins the semantic world by linking to a service agent The non-agent program can be wrapped with a service agent shell Trustworthy: Owned by the semantic web site May 6, 2004 2

Function Level I. II. Provide the requested semantic web page Answers simple questions about the semantic web pages: The inference in this level is based on local rules, limited semantic pages and local ontologies III. Answer complicated questions about the semantic web pages: The inference in this level involves multiple ontologies, multiple semantic web sites and multiple agents IV. Validates trust and delegation May 6, 2004 3

OWL as a Semantic Language n n n Well-defined model-theoretic semantics Unambiguously computer-interpretable Facilitates a higher-level of interoperability between the agents By agreeing on how meaning is conveyed, applications can share meaningful content easily and naturally May 6, 2004 4

OWL as ACL Content Language I. II. IV. V. OWL’s expressive power is adequate for many needs of current agent based systems OWL offers better support for using terms drawn from multiple namespaces and multiple ontologies than existing ACL content languages OWL provides improved support in modeling, maintaining and sharing ontologies OWL is designed to fit into and integrate with webbased information and web services OWL has the potential to be a widely accepted and used representation language, enhancing the potential for interoperability among many systems May 6, 2004 5

Semantic Web in FIPA n n n FIPA is the most widely used MAS framework n Well developed and documents standards n Good open source software RDF is one of FIPA’s standard content languages OWL is widely used for publishing ontologies within the FIPA community, for example, agentcities and open. Net May 6, 2004 6

FIPA Standards Overview IDL Envelope. Encoding. Scheme XML Envelope Owl for publishing bit-eff 1 communicative acts 1 is. Transmitted. Over 1 IIOP HTTP contains String ACLEncoding. Scheme XML bit-eff is. Expressed. In 1. . * ACL Message String CLEncoding. Scheme Owl for publishing request, query, request-when protocols contract-net, iterated-contract-net contains 1 Content is. Expressed. In 11 1 contains 0. . * Interaction. Protocol 1. . * May 6, 2004 ACL Content SL Language for Owl ontologies fipa-agent-management belongs. To Symbol --Tim Finin 2003 1 Owl as a content language 1 1 brokering, recruiting subscribe, propose Transport Protocol Ontology 1 7

Owl for user models and profiles Owl for representation and reasoning FIPA Agent Platform software Agents belong to one or more agent platforms Owl for authorization which provide service basic policies descriptions services. AA AMS DF ACC IIOP internal platform message transport --Tim Finin 2003 May 6, 2004 8

Outline Part 1: Thesis and Contribution Part 2: Background Part 3: Research Question Part 4: Agent-Based Services Part 5: TAGA and F-OWL Part 6: Conclusion May 6, 2004 9

Why TAGA ? n n n Need a big and complicated system to evaluate the ideas Agentcities provides a robust global agent services platform TAC is a successful travel market simulation system May 6, 2004 10

Trading Agent Competition n n The Trading Agent Competition proposed (1999) and first run (2000) by Michael Wellman and Peter Wurman Goal: promote and encourage research in markets involving autonomous trading agents; Methods: trading agents operate within a travel market scenario; International competitions in 2000, 2001 and 2002 were based on a simple travel scenario May 6, 2004 11

Problems n n n TAC classic assumes that agents interact via a few centralized markets. Technology is basically client-server with well defined APIs and simple XML encoding. Real word interactions are varied and rich n n n Customers can chose to interact with travel agencies, dynamic markets, or directly with service providers Choices are governed by value, speed, reputation Rich information exchange abounds -- customers have complex interests and preferences, service providers have detailed descriptions, etc. Common ontologies are important Trust and reputations are important. May 6, 2004 12

TAGA Features • • • Open Market Framework Auction Services OWL Message Content Travel Market Ontology Global Agent Community • Goal: test bed for experimenting with Agents, Semantic Web and Web Services May 6, 2004 13

A Typical Scenario 4 Airline WS Hotel WS 6 Market Oversight Agent 3 Bulletin Board 1 Auction Service 2 a 2 b CA TA 5 May 6, 2004 14

TAGA Agents (1) One CA joins the Game every 30 Sec. …. v Find travel arrangements v Save $$ Customer Agents v Organize travel v Maximize profits TA-1 (AAP) TA-2 (AAP) TA-4 (JADE) Entertainment Airline Hotel Web Service May 6, 2004 v sell “goods” v Maximize profits 15

TAGA Agents (2) v Helps CA find one or more TA Bulletin Board Agent v Operates the auctions markets: English, Dutch, Priceline and Hotwire. Auction Service Agent Market Oversight Agent v Manage the financial records v Announces the winning TA May 6, 2004 16

Dynamic Contract Protocol May 6, 2004 17

Priceline Auction Protocol May 6, 2004 18

Technology n System Infrastructure: n n Travel and Auction Ontologies: OWL Web Services: WSDL Web technology: n n n Agentcities + AAP + JADE Apache, My. SQL Java Web Start Agent Communication: n n FIPA (OWL as the content language) Service registration: OWL-S May 6, 2004 19

Simulation Design n n Game running continuously Travel Agent n n Direct Buy or Bid Acquire resource before or after win customer Penalty, reputation Auction Service Agent n n English & Dutch auction “Name your price” auction (priceline. com & hotwire. com) May 6, 2004 20

Travel Agent Game in Agentcities Motivation Features Market dynamics Auction theory (TAC) Semantic web Owl for Agent collaboration (FIPA & Agentcities) negotiation Technologies Open Market Framework Auction Services OWL message content Owl for OWL Ontologies publishing Global Agent Community Ontologies FIPA (JADE, April Agent Platform) Semantic Web (RDF, OWL) Owl for Web (SOAP, WSDL, DAML-S) contract Internet (Java Web Start ) http: //taga. umbc. edu/ontologies/ communicative acts travel. owl – travel concepts fipaowl. owl – FIPA content lang. auction. owl – auction services tagaql. owl – query language enforcement Report Direct Buy Transactions Report Contract Owl for representation and reasoning Report Auction Transactions Market Oversight Agent Customer Agent Proposal P Report Travel Package Owl for modeling Agents Travel trust d Bi id R Bulletin Board Agent CF B eq t es u Auction Service Agent Direct Buy Owl for service descriptions Web Service Owl as a Agents content FIPA platform infrastructure services, including directory facilitators enhanced to use OWL-S for service discovery language http: //taga. umbc. edu/

ACL Content n n Statements: the price of this hotel in day 3 is $100/night; Requests: create an airline auction instance; Contracts: if the Travel Agent TA 1 successful organized the travel package, customer Joe will pay $400 to TA 1, else, TA 1 pay $200 compensation to Joe. Policies: to win the contract of the customer Joe, the travel agent must have reputation better than average May 6, 2004 22

Ontologies ACL http: //taga. umbc. edu/ontologies/fipaowl. owl n Auction http: //taga. umbc. edu/ontologies/auction. owl n Travel http: //taga. umbc. edu/ontologies/travel. owl n Submitted to DAML ontology library May 6, 2004 23

Multiple Ontologies Support n n New. Instance Ontology. Query Ontology. Share Ontology. Relation Agent A New. Instance Agent B Ontology. Query Ontology. Share Ontology. Query Agent B Ontology. Relation May 6, 2004 24

TAGA in Action TAGE Home Page http: //taga. umbc. edu Create a TAGA game online Download the latest TAGA pkg and docs http: //taga. umbc. edu/taga/download/ TAGA on Agentcities network (UMBCTac. agentcities. net) Baltimore, MD USA http: //www. agentcities. net/ TAGA supports heterogeneous agent platform. A FIPA-JADE agent can interact with a FIPA-AAP agent May 6, 2004 25

Conclusion (1) n n n A rich framework for exploring agent-based approaches to e-commerce applications. Auction services are developed to enrich the Agentcities environment The use of Semantic Web languages (OWL) improves agent interoperability OWL-S is employed to support agent service registration, discovery and invocation A sourceforge project May 6, 2004 26

Conclusion (2) n n Won the Best Student Entry in the Agentcities sponsored Agent Technology Competition held at Barcelona in Feb. 2003 TAGA platforms have been running in Agentcity. Net for more than 16 months. Invited to Intelligent System Demonstrations at IJCAI 2003 and AAMAS 2004 Used by people from US, Korea, Romania, etc. May 6, 2004 27