250 likes | 392 Views
Multiagent systems and E-Commerce. 제조통합자동화 연구실 세미나 발표자 : 정성원 발표일자 : 2001.3.23. Reference Papers. Software Agents Michael R. Genesereth COMMUNICATIONS OF THE ACM VOL 37 , NO 7, 1994 Cooperative Multiagent Systems: A Personal View of the state of the Art Victor R. Lesser
E N D
Multiagent systems and E-Commerce 제조통합자동화 연구실 세미나 발표자 : 정성원 발표일자 : 2001.3.23
Reference Papers • Software Agents • Michael R. Genesereth • COMMUNICATIONS OF THE ACM VOL 37 , NO 7, 1994 • Cooperative Multiagent Systems: A Personal View of the state of the Art • Victor R. Lesser • IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING VOL 11, NO 1, 1999
Contents • Software Agent • Cooperative Multiagent Systems • A Brokering Protocol for Agent-Based E-Commerce • Discussion
Agent • What is Agent? • It perceives the world in which it is situated. • It has the capability of interacting with other agents. • It is pro-active in the sense that it may take the initiative and persistently pursues its own goals.
Software Agent • What is software agent? • A software agent is a program that performs a specific task on behalf of a user, independently or with little guidance • A software agent exchanges the data and logical information , individual commands and scripts by its common language • Questions • What is an appropriate agent communication language? • How do we build agents capable of communicating in this language? • What communication architectures are conductive to cooperation?
Agent Communication Language • A universal communication language • Procedural approach • Based on the idea that communication can be best modeled as the exchange of procedural directives • Scripting language – TCL • Declarative approach • Based on the idea that communication can be best modeled as the exchange of declarative statements • ACL • The vocabulary ACL • Inner language called KIF(Knowledge Interchange Format) • Outer language called KQML(Knowledge Query and Manipulation Language)
KIF(Knowledge Interchange Format) • What is KIF? • KIF, a particular logic language, has been proposed as a standard to use to describe things within computer systems, e.g expert systems, databases, agents, etc. • A prefix version of first order predicate calculus, with various extensions to enhance its expressiveness • Examples • <salary 015-46-3946 widgets 72000> • (>(*(width chip1)(length chip1))(*(width chip1)(length chip1))) • (=>(and (real-number ?x) (even-number?n))(> (expt?x?n) 0))
Performative (ask-one : content (PRICE IBM ? Price) : receiver stock-server : language LPROLOG : ontology LYSE-TICKS) : Message type KQML (Knowledge Query and Manipulation Language) • What is KQML • KQML is a Lisp-based language that was developed as part of the ARPA Knowledge Sharing Effort. • KQML is based on speech act theory, and message types are indicated by performative. • Example
Predefined performatives of KQML • Basic informatives (constatives) • Tell (share a piece of knowledge) • Deny (retract or negate a speech act) • Untell (retract a statement ; equals deny tell) • Query performatives • Evaluate (evaluate an expression ; details depend on language) • Reply (I am sending you data to answer your query) • Ask-if (Yes-no question) • Ask-one (send me one response that matches my query) • Ask-all (send me all responses that match my query) Predefined performatives of KQML 1993
KQML performative parameters : sender symbol identifying the sender : receiver symbol identifying the recipient : reply-with identifier that must appear in the reply : content the content of the message : language language in which content is expressed : ontology ontology used by content • Some performatives take additional parameters • There are defaults for parameters that are omitted Basic set of KQML performative parameters 1993
Multiagent system • What is Multiagent System? • A collection of, possibly heterogeneous, computational entities, having their own problem solving capabilities and which are able to interact in order to reach an overall goal. • MAS is seen as a system revealing a kind of synergy that would not be expected from the sum of its component agent.
Application of Multiagent Systems • Applications Domain (Examples) • Distributed situation assessment • Distributed resource scheduling and planning • Distributed expert systems • Advantages • Speed-up due to concurrent processing • Less communication bandwidth requirements because processing is located nearer the source of information • More reliability because of the lack of a single point of failure • Easier system development due to modularity coming from the decomposition into semiautonomous agents
The Nature of Multiagent Interaction • The need to interact when agents solve the subproblems • The subproblems are the same or overlapping • The information one agent has is needed to solve subproblems • The interactions among agents in a multiagent system • Impossible solve subproblem pi without solving pj • Solving pj may simply make it easier to solve pi • Knowing the solution to pj may obviate the need to solve pi
Key Principles Used In Building Multiagent Systems • The need to view the performance of a multiagent system in terms of an interdependent set of criteria • There is no single approach to organizing agent behavior that will be right for all situations • The need for flexibility in agent problem solving • Sophisticated domain problem-solving architecture • Increasing the scope of activities • The need to exploit the efficiencies of organized behavior in coordinating large agent societies • Organizing the agents in terms of roles and responsibilities can significantly decrease the computational burden on coordination decisions
A Brokering Protocol for Agent-Based E-Commerce Kwang Mong, Sim and Raymond Chan IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS VOL 30, NO 4, 2000
Introduction(1/2) • Some of activities of information brokering in e-commerce • Connecting buyers and sellers • Selection, evaluation, filtering and assignment Request Advertisement
Introduction(2/2) • The goals of this research • Design and implement an algorithm that connects buyer and seller agents • Devise a brokering protocol for specifying and structuring the interactions among electronic intermediaries and trading agents • Design and engineer a testbed that models and simulates some of the activities of information brokering in the domain of securities trading
Connecting Buyers and Seller(1/4) • Selecting request and advertisements • That are not expired or withdrawn • With profiles and preferences that are closely matched • That have not been previously assigned
Connecting Buyers and Seller(2/4) • Evaluating Connections • The utility U • A domain specific attribute <a1,a2, ….. ,an> • The utility of a connection Ci • For Example – Securities trading • Price(P), Volume(V) and Desirability(D) • Values of P,V and D range from 0 to 1 with the following interpretations
Connecting Buyers and Seller(3/4) • Filtering • Cutoff point • A smaller value of , A larger value of • The need to optimizing value of
Connecting Buyers and Seller(4/4) • Assigning buyers and seller • This stage makes the completion of the selecting cycle for a request and releases the connection results to buyers and sellers for review • Stopping condition • Number of connection equals to the expected number of response specified by the buyer • Request has already gone through a prespecified number of cycles of selection and filtering
Recommend Agent-Based Information Brokering Testbed Agent-based information brokering testbed The stages of brokering protocol
Information Exchange Among Agents • The types of messages • Requests from buyer agents • Advertisements from seller agents • Connections of request to advertisements • Withdrawals of requests or advertisements • Results of decisions Ex) Request ( request-id : id-type; buy-agent-id : id-type; buy-stock-name : stock-name; buy-valid-period : period; buy-volume : volume; buy-price : double; expected-response : integer; )
Experimentation and Evaluation • Experiments • Average Connection Time – stable • Satisfactions • Balanced Loading
Discussion • What can we do with agents? • Analysis system and make a framework • Define the type of message • ……. • How can we validate our framework?