Cooperative Meeting Scheduling among Agents based on Multiple ...

Cooperative Meeting Scheduling among Agents based on Multiple ...

Cooperative Meeting Scheduling among Agents based on Multiple Negotiations Toramatsu SHINTANI and Takayuki ITO Department of Intelligence and Computer Science, Nagoya Institute of Technology [email protected] JAPAN Motivation Distributed Meeting Scheduler Distributed scheduling Implementation Reaching a Consensus Multiple Negotiations (Persuasion Process) Preference Revision Using Private Preferences Conclusions Motivation In social decision making A trade-off between "reaching a consensus" and "maximizing own expected payoffs (private preferences)" Membership in a coalition may maximize a personal outcome. Some of solutions can be based on "Settoku" (persuasion) in Japanese social decision making. Realizing an architecture for a mult-agent negotiation in distributed scheduling Background [Sen and Durfee, 1998]

Using a central host agent User preferences are not taken into account [Haynes,et al., 1997] User's preference by values with a thread Adjusting values under a threshhold which means a dgree to which a value can or cannot be compromised [Garrido and Sycara, 1996] Using user's preferences with high joint utility They did not establish how to reach an agreement among agents and compromise with other agents The aim is to propose a new architecture for multiagent- negotiation in a distributed meeting scheduler based on the persuasion mechanism by using user's preference. Distributed Scheduling System December December December 1996 1996 1996 S M SS M MT TT W W WT TT F

FF S SS December 1996 Agent Calendar An agent 1 11 2 22 S3 33 4 44 5 55 6 66M7 77 8 88 9 991011121314 1011121314 1011121314 1 2 15161718192021 15161718192021 15161718192021 22232425262728 22232425262728 22232425262728 293031 293031 293031

is assigned to an user. Agents negotiate using the private data. T W T F S 3 4 5 6 7 10 11 12 13

14 8 9 15 16 17 18 19 20 21 22 23 24 25 26 27

28 29 30 31 The Calendar 1 6 0 24 5 2 An 1: July Not Event 2001 important, (a time interval) Slightlyprivate important, December 1996 12/7/2001-9:00-12:00-weight(5) A

Atime calendar slot is used for3:keeping schedules of an user. 5: Strongly important, 7: Very strongly important, The schedule includes date, hours, and events given weights. 9: Extremely important For convenience, we can use the verbal scale for putting a weight of an event. S M T W T F S 1 2

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22 23 24 25 26 27 28 29 30 31 A time slot The Distributed Meeting Scheduling Request for

a meeting Deciding attributes of the meeting and designing alternatives Quantifying the user's preference based on MAUT Getting a Result Negotiation among agents Negotiation among agents Negotiation among agents The Multiple Negotiations By using the persuasion process, agents negotiate with other agents using users' private preferences of the meeting. In this phase, the multiple negotiations are conducted by agents. The Outline of the Persuasion Process Persuasion between agent A and agent B. We call A "Persuader" and B "Compromiser." proposal A Agreement Can I accept?

B 1. Proposal 2. Preference revision 3. Agreement 1. A sends a proposal to B. 2. B tries to revise her preference. 3. If B could revise her preference, they reach an agreement. Agent A A1A2A3 Agent B A A 2 1 A3 persuade The Multiple Negotiations host clone 2. Reporting 1. Cloning 3.

Exchanging Persuasion Pattern Multiple Each 4321agent negotiationsher dispatches results information clones Quantifying User's Preference Using Multiple Attribute Utility Theory Scheduling a meeting preference Convenience 9:00-10:00 Size Length 9:00-11:00 13:00-14:00 We can select several options with respect to f according to the application area. In our system, we select the AHP method for calculating user's utility.

Quantifying User's Preference Using AHP AHP The pairwise comparison matrix with respect to the criterion "Convenience" 9:00-10:00 9:00-11:00 13:00-14:00 Scheduling a meeting Weights 9:00-10:00 1 1/3 2 0.205 9:00-11:00 3 1 9

0.705 13:00-14:00 1/2 1/9 1 0.089 Convenience 9:00-10:00 Size Length 9:00-11:00 13:00-14:00 The Preference Revision In the preferece revision, agents try to change the weights of alternatives. In order to change the weights, agents try to adjust the weights of criteria within 2 intervals. The fuzzy measurement enables agents to adjust the weights of criteria. Slightly

E 6 9 1 2 4 5 7 8 3 Very Equally Strongly xtremely Strongly 2 intervals INPUT : The persuaderfs most preferable alternative( PA) and The Preferece Revision the compromiser fs original preference( PreF) Algorithm OUTPUT: Success or Failure Function

PrefRevision (PA,Pref) PATS :=apowersetofattributesforthealternativeP ; A. SortedPATS :=sortBySize (PATS); Candidates :=; Solutions :=; For each ATS inSortedPATS ATS :=IncreaseV alues (ATS); Pref :=ReCalculate (Pref,ATS); Candidates :=Candidates Pref; If PA ==theMostP referableAlternative (Pref ) Then Solutions :=Solusions Pref; If Solutions is not empty Then Pref:=selectMinimalPref (Solusions ); return Success

PATS := a power set of attributes for the Most preferable MA); alternative( SortedPATS :=sortBySize (PATS); For each CandidateP ref inCandidates For each ATS inSortedPATS ATS :=decreaseV alue (ATS); Pref :=ReCalculate (CandidateP ref,ATS ); If PA ==theMostP referableAlternative (Pref ) Then Solutions :=Solusions Pref; If Solusions is not empty Then Pref :=selectMinimalPref (Solusions ); return Success return Failure

End Function The Feature of the Preference Revision The MC principle An agent should change an user's preference as minimal as possible The OC principle An agent should change an user's preference based on the preference order of alternatives In our system, a compromiser tries to adjust attribute values based on " generate and test" style. The problem is that the solution space is too huge to revise agent's preference. Implementation December 1996 December December 1996 1996 M TT T W W TT T FF F SS S SS S M M W

on Distributed MacOS,Unix, Scheduling and Windows system MiLog: A Mobile Agent Framework 11 1 22 2 33 3 44 4 55 5 66 6 77 7 1011121314 88 8 99 91011121314 1011121314 15161718192021 15161718192021 15161718192021 22232425262728 22232425262728 22232425262728 293031

293031 293031 The Main Features of MiLog The cloning Hybrid programming technique for mobile by usingagents Java and enables prolog us to realize Java concurrent API fornegotiation Java Progamming processes for the multipleProlog negotiation. Predicates for Prolog Programming Realizing mobile agents Strong Migration clone Multi-thread programming Suspend/Resume/Interrupt WWW server functionality web-service/access functions package sample; import java.util.*; import millog.*;

public class Sample { Milog milogAgent; public Vector complexReturnValue() { if( milogAgent.syncQuery("append([abc],[def,ghi],[X|Y]).") != null ) { String answer1 = (String) milogAgent.getAnswerAsObjects("X"); Vector answer2 = (Vector) milogAgent.getAnswerAsObjects("Y"); return(answer2); } return(null); } Java programming with MiLog Agent An Example of MiLog WWW Agent GUI Program for Monitor Inspector Agent (Prolog) Inerface Experimental Result Conclusions

A new multi-agent negotiation The multiple negotiations can reflect user's individual preferences. The preference revision effectively find a solution for a compromiser in the persuasion process. The Distributed Meeting Scheduler realizing a cooperative meeting scheduling among agent improving a trade-off between "reaching a consensus" and "reflecting users' preference" in a social decision. The result shows that the multi-agent negotiation based on private preference is an effective method for a distributed meeting scheduler. The process can facilitate reaching a consensus among agents.

Recently Viewed Presentations

  • NACG PPT Template

    NACG PPT Template

    - Typically, light load requirements (25 kN to 200 kN), in Alberta, Saskatchewan and . Manitoba - Higher load requirements in British Columbia (< 1,000 kN) - Commonly without permanent steel casing. Absence of Steel Casing? - Rarely is criteria...
  • Chapter 6 -Congress

    Chapter 6 -Congress

    Types of Bills. Two Types of Bills. Private Bills: concern individual people or places . Public bills: apply to the entire nation and general matters like taxation, etc. Congress considers many . resolutions (formal statements of opinions from lawmakers) Joint...
  • Percentages Questions and Answers - chsmaths

    Percentages Questions and Answers - chsmaths

    Percentages Questions and Answers. Fractions, Decimals and Percentages. Finding Percentages. Percentage Increase/Decrease. Reverse Percentages. You tube playlist LINK
  • Co ll eg ea nd A s It

    Co ll eg ea nd A s It

    Students are allowed to put as much money as they choose in each jar and the teacher with the most money after the set amount of time has to kiss a pig Spirit Chairs Students who bring registration in on...
  • Antibiotic Cycling A solution to the evolutionary arms

    Antibiotic Cycling A solution to the evolutionary arms

    Antibiotic Cycling A solution to the evolutionary arms race? Presented by Alexandra Roland An "evolutionary arms race" is an evolutionary struggle between competing sets of co-evolving genes that develop adaptations and counter-adaptations against each other, resembling an arms race.
  • Wildlife Management and Vector Control for an FAD

    Wildlife Management and Vector Control for an FAD

    In an FAD outbreak, susceptible wild animal species may become infected, serve as a reservoir, or further spread the pathogen to domestic livestock or poultry. Terms describing these modes of involvement as used in this presentation are defined here. Feral:...
  • November 1, 2016 Objective: To create a shareable

    November 1, 2016 Objective: To create a shareable

    Identify and define irony, allusion, and point of view. Identify and define a verb. Distinguish between action verbs, linking verbs, and helping verbs. Discuss theme and how it is developed throughout a work. Write a narrative. Write a text-based analysis...
  • Energy and the Environment

    Energy and the Environment

    Energy and the Environment Environmental Problem Sources Philosphy Lifestyles that emphasize consumption Consumption vs. Quality Yugo vs. Mercedes Disposal ball point pens Fast food packaging Environmental Problem Sources Combustion & air pollutants Incomplete combustion Carbon monoxide Volatile hydrocarbons Soot and...