1 / 21

Welfare Engineering in Practice: On the Variety of Multiagent Resource Allocation Problems

Welfare Engineering in Practice: On the Variety of Multiagent Resource Allocation Problems. Yann Chevaleyre 1 , Ulle Endriss 2 , Sylvia Estivie 1 and Nicolas Maudet 1. (1)LAMSADE, Univ. Paris IX-Dauphine (2)Dept. of Computing, Imperial College London. Introduction.

tacy
Download Presentation

Welfare Engineering in Practice: On the Variety of Multiagent Resource Allocation Problems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Welfare Engineering in Practice: On the Variety of Multiagent Resource Allocation Problems Yann Chevaleyre1, Ulle Endriss2, Sylvia Estivie1 and Nicolas Maudet1 (1)LAMSADE, Univ. Paris IX-Dauphine (2)Dept. of Computing, Imperial College London

  2. Introduction • Recurring problems like E-auctions, patrol … • Similarities between these problems ? Not exploited yet… • Formalize this similarities for a category of problem : Resource allocation problem • Why??? • A lot of theoretical result for resource allocation • Possibility to develop a platform

  3. Talk Overview • Welfare Engineering • Designer scope • Resource Allocation Framework • Example Applications • Criteria • Conclusion

  4. Welfare Engineering How we can make agents negotiate socially optimal outcomes? • Social welfare ordering (quality of the solution) • Social interaction mechanism (to arrive at a solution) • Behaviour profiles (interaction mechanism) Socially optimal allocation of resources

  5. Talk Overview • Welfare Engineering • Designer scope • Resource Allocation Framework • Example Applications • Criteria • Conclusion

  6. The Problem of the Designer Scope Which agent does designer control? • [Wurman et al 02] • Agent scope • Mechanism scope • System scope • Proprietor role • End-user role Agent

  7. Talk Overview • Welfare Engineering • Designer scope • Resource Allocation Framework • Example Applications • Criteria • Conclusion

  8. u2(A) u1(A) 2 1 R A u4(A) u3(A) 4 3 A Resource Allocation by Negotiation • Finite set of agents A and finite set of discrete resourcesR • An allocationA is a partitioning of R amongst the agents in A • Every agent i Ahas a utility function ui(A)

  9. Social Welfare Majoring the well being of a society Social welfare is tied to the welfare of a society’s weakest member • Egalitarian social welfare • Utilitarian social welfare • Anything that increases average utility • is taken to be socially beneficial • Envy-freeness social welfare • There is zero probability of having an • agent envying somebody else • Research issue : the impact of individual utility on social welfare

  10. Our framework (1/2) • Monetary payments • Deal couple with monetary side payment • Payment function • Limited money • Approximating flows • Representation of continuous resources (water, energy, …)

  11. Our framework (2/2) • Roles • Sellers • Buyers • … • Protocol restrictions • Restrictions on the negotiation protocol

  12. Talk Overview • Welfare Engineering • Designer scope • Resource Allocation Framework • Example Applications • Criteria • Conclusion

  13. ? ? Examples of Applications (1/3) • Multiagent Patrolling (1/2) • The multiagent patrolling problem: how should agents move around an area such that every part of the area is visited the most often ? • Goal : find strategies which minimize the time between 2 visit on each node

  14. Examples of Applications (1/3) • Multiagent Patrolling (2/2) • Multiagent patrolling applies to: • Multi-robot applications (intrusion detection, cleaning team of robots, delivery) • Video-games (in warcraft-like games, doom-like, …) • Military application (surveillance, tracking intruders) • Internet applications • Resources : each node • Utility of each agent : how well it patrols over the node it owns • Resource allocation : agent can exchange nodes in order to maximize his patrolling performance

  15. Examples of Applications (2/3) • Allocation of satellite resources [Lemaitre et al 03] Resources initially held by the virtual proprietor Agents send observation request

  16. Examples of Applications (3/3) • E-Auctions • Different kinds of e-auction • B2C (Business to Consumer) : antique dealer • C2C (Consumer to Consumer) : eBay like • B2B (Business to Business) : FCC, fairmarket… • Similarities and differences : but all could be represented with a model of resource allocation. • Roles : sellers and buyers

  17. Talk Overview • Welfare Engineering • Designer scope • Resource Allocation Framework • Example Applications • Criteria • Conclusion

  18. Criteria for a Social Welfare Selection (1/2) Proprietor gain • Utility-dependent • Example : tax on gain • Example of application uses it : Multiagent Patrolling • Transaction-dependent • Example : tax on each transaction • Example of application uses it : e-auctions • Membership-dependent • Example : Entrance fees • Example of application uses it : Satellite allocation, e-auctions

  19. Criteria for a Social Welfare Selection (2/2) Application dynamics Between a run • Possibility for an application to run several times • Yes : Satellite application, C2C e-auctions • No : FCC e-auctions • If yes, whether and how the characteristics could be modified between runs? • C2C e-auctions : users may be different

  20. Conclusion • Multiagent resource allocation : A powerful paradigm • The first idea of social welfare choice in not necessarily the better.[Guttman, Maes 99] Toward a test platform

  21. References • [Guttman, Maes 99] R.H. Guttman and P. Maes. Agent Mediated integrative negotiation for retail electronic commerce. In Agent Mediated Electronic Commerce, 1999. • [Lemaitre et al 03] M. Lemaitre, G. Verfaillie, H. Fargier, J. Lang, N. Bataille and J.M. Lachiver. Equitable allocation of earth observing satellites resources. In Proc of the 5th ONERA-DLR Aerospace Symposium (ODAS’03), 2003. • [Wurman et al 02] P.R. Wurman, M.P. Wellman, and W.E. Walsh. Specifing rules for electronic auctions. AI Magazine, 23(3):15-23, 2002.

More Related