Model-driven Specification of Strategies for Negotiating Agents Ren - PowerPoint PPT Presentation

Institute of Business Information Systems Model-driven Specification of Strategies for Negotiating Agents René Schumann , Zijad Kurtanovic, and Ingo J. Timm Valencia, May 4 th 2012

Institute of Business Information Systems Outline • Motivation • Meta Model for Tradeoff Strategies • Transformation • Example • Conclusion

Institute of Business Information Systems Motivation • Agents should negotiate on behalf of humans • Responsibility remains on the humans • Outcomes have to be predictable – By guarantees – Specify the strategies of the agents • Mismatch between capability and responsibility (programmer vs. manager)

Institute of Business Information Systems Idea • Allow managers to specify the negotiation strategy for “its” agent. • Using MDD techniques to automate the translation from manager to agent (based on the ECore meta model) • Focus : Tradeoff strategies – A set of tradeoff relations which defines a preference between two assignments of negotiation attributes

Institute of Business Information Systems The Meta-Model

Institute of Business Information Systems The Meta-Model

Institute of Business Information Systems Graphical Representation • For the most important concepts graphical representations have been defined e.g.: – Negotiation Strategy <Service Name>

Institute of Business Information Systems Graphical Representation • For the most important concepts graphical representations have been defined e.g.: – Negotiation Strategy <Service Name> Price – Negotiation Attribute [120,270]

Institute of Business Information Systems Graphical Representation • For the most important concepts graphical representations have been defined e.g.: – Negotiation Strategy <Service Name> Price – Negotiation Attribute [120,270] – Trade-off relation

Institute of Business Information Systems Graphical Representation • For the most important concepts graphical representations have been defined e.g.: – Negotiation Strategy <Service Name> Price – Negotiation Attribute [120,270] – Trade-off relation • But currently no graphical editor is implemented

Institute of Business Information Systems How to edit it now (Eclipse Widget)

Institute of Business Information Systems Case Study: Scenario • Bilateral Negotiation – Multiple parameters – preference over outcome – Minimal information disclosure – Minimal relaxing of desires

Institute of Business Information Systems Strategies Buyer’s strategy

Institute of Business Information Systems <Service Name> Strategies (10) (7) Buyer’s strategy Price Accounting period [120,270] {1,3,6} (230,2) (200,18) Timeliness Contract duration {1,2,3,4,5,6} {6,12,18,24} Seller’s strategy

Institute of Business Information Systems Transformation • Use a relational representation – Computation can be done at compile time – Fast at run time (negotiation time) Algorithm 1 Pseudo code of the tree transformation algorithm function generateTables (set of trees forming a tradeo ff strategy) for each Tree do generate a representation set of the root Call InduceRepresentationSet(root) end for end function function InduceRepresentationSet (Node X) if X NOT ROOT then Induce a representation set from X end if for each Child-Node C of X do InduceRepresentationSet(C) end for end function

Institute of Business Information Systems Transformed Strategies <Service Name> (10) (7) Price Accounting period [120,270] {1,3,6} (230,2) (200,18) Timeliness Contract duration {1,2,3,4,5,6} {6,12,18,24}

Institute of Business Performative: Find Buyer Information Systems Constraint: PR ≤ 160 Negotiation Round 1 Performative: Check Seller (PR:150,AC:4,CD:18,AP:1) Performative: Find Buyer Round 2 Constraint: PR ≤ 160 ∧ AC ≤ 3 Seller Performative: Relax Performative: Find Buyer Constraint: PR ≤ 145 ∧ AC ≤ 3 Round 3 Seller Performative: Relax Performative: Find Buyer Constraint: PR ≤ 175 ∧ AC ≤ 3 Round 4 Performative: Check Seller (PR:165,AC:2,CD:18,AP:1) Performative: Find Buyer Round 5 Constraint: PR ≤ 175 ∧ AC ≤ 3 ∧ CD ≤ 13 Seller Performative: Relax Performative: Find Buyer Constraint: PR ≤ 190 ∧ AC ≤ 3 ∧ CD ≤ 15 Round 6 Performative: Check Seller (PR:180,AC:2,CD:12,AP:1) Performative: Find Buyer Constraint: PR ≤ 190 ∧ AC ≤ 3 ∧ CD ≤ 15 ∧ AP ≥ 2 Round 7 Performative: Check Seller (PR:180,AC:2,CD:12,AP:3) Round 8 Buyer Performative: Deal Table 1. Full negotiation trace of buyer and seller (PR: price, AC actuality, CD c

Institute of Business Information Systems Conclusion • MDD approach to close the gap between responsibility and capability for agent-based negotiations • Bottom-up approach that can be integrated in other MDD approaches (top-down) • Vision: Managers can specify strategies by themselves.

Institute of Business Information Systems Questions? Contact ¡ ¡René ¡Schumann ¡(rene.schumann@hevs.ch) ¡