Coevolutionary Computation Based Iterative Multi-Attribute Auctions

Abstract

Multi-attribute auctions extend traditional auction settings. In addition to price, multi-attribute auctions allow negotiation over non-price attributes such as quality, terms-ofdelivery, and promise to improve market efficiency. Multi-attribute auctions are central to B2B markets, enterprise procurement activity and negotiation in multi-agent system. A novel iterative multi-attribute auction mechanism for reverse auction settings with one buyer and many sellers is proposed based on competitive equilibrium. The auctions support incremental preference elicitation and revelation for both the buyer and the sellers. Coevolutionary computation method is incorporated into the mechanism to support economic learning and strategies for the sellers. The myopic best-response strategy provided by it is in equilibrium for sellers assuming a truthful buyer strategy. Moreover, the auction are nearly efficient. Experimental results show that the coevolutionary computation based iterative multi-attribute auction is a practical and nearly efficient mechanism. The proposed mechanism and framework can be realized as a multi-agent based software system to support supplier selection decision and/or deal decision for both the buyer and the suppliers in B2B markets and supply chain.