Auctions for robotics panel: talking points Robotics setting –Incentives usually don’t matter –Problems are combinatorial/multi-attribute => modern work on complex auctions & exchanges can be helpful –Similar to other MAS (at least from a coordination perspective)
Peer-to-peer negotiation Marginal cost based contracting [AAAI-93, ICMAS-95, PhD-96] –Automated cost computation –Issues emerging from distributed implementation Parellellism vs monotonicity Avoiding msg saturation Termination, … –Contracting as hill-climbing [AAAI SS-98] –OCSM-contracts [AAAI SS-98, ICMAS-98, AAAI-99, ICDCS-00] Leveled commitment contracts [ICMAS-95, AAAI-96, IJCAI-99, GEB-01, AIJ-02] –Sequences – cascades [ICMAS-98, J. Econ. Dynamics & Control-01]
Mediated markets Removes negotiation process uncertainty => better allocations Usually faster as well Package bidding expressive competition [DCR-01, GEB- 06, Interfaces-06, IAAI-06, …] –Rich forms of offer constructs –Side constraints –Multi-attribute functionality Preference elicitation from the different parties (studied for CAs & CEs already) [EC-01, AAAI-02, EC-03, …] –Focuses the agents’ marginal cost/value computations
Deliberation control Heuristically in peer-to-peer negotiation [AAAI-93, ICMAS-95, PhD-96] Game-theoretically –Impossibility results [ICMAS-96, ICEC-00, AAMAS-05] –Using performance profile trees in auctions [TARK-01, AGENTS WS-01, AAMAS-03, AAMAS-04] in bargaining [AIJ-01, AAMAS-02]
Online problem Has been studied for multi-unit –auctions [Lavi & Nisan EC-00, …] –exchanges [Blum, Sandholm, Zinkevich SODA-02, JACM-06] Thank you for your attention!
Preference elicitation from multiple agents
Monsters Local planning complexity Communication complexity (Loss of privacy)
Clearing algorithm What info is needed from an agent depends on what others have revealed Elicitor Conen & Sandholm IJCAI-01 workshop on Econ. Agents, Models & Mechanisms, ACMEC-01 Elicitor decides what to ask next based on answers it has received so far $ 1,000 for $ 1,500 for ? for