Bid Determination in Simultaneous Auctions An Agent Architecture

Suppose you want to buy a used Canon AE-1 SLR camera and ash at an on-line auction. At last count, over 4000 links to on-line auction sites were available at advocacy-net.com. It would be quite a daunting task to manually monitor prices and make bidding decisions at all sites currently o ering the camera|especially if the ash accessory is sometimes bundled with the camera, and sometimes auctioned separately. But for the next generation of automated trading agents, this will be a routine task. Simultaneous auctions, or parallel auctions, present a challenge to bidders, particularly when complementary and substitutable goods are on o er. Complementary goods are items such as a ash, a tripod, and a case, that would complement a camera|but a bidder desires any of the former only if s/he is certain to acquire the latter. Substitutable goods are goods such as the Canon AE-1 and the Canon A-1|a bidder desires one or the other, but not both. In contrast to combinatorial auctions, in which bids may be placed for combinations of items (e.g., \camera and ash for $295"), simultaneous auctions require separate bids to be placed for each individual item. In combinatorial auctions, the problem of choosing a set of winning bids that maximizes revenue|the so-called winner determination problem (WD)|falls in the hands of the auctioneer; in simultaneous auctions, the complexity burden lies with the bidders. In this paper, we present an agent architecture for intelligent bidding in simultaneous auctions. Our architecture is centered on a class of problems we call bid determination (BD): determining what bids to place in simultaneous auctions, given current market conditions and utilities on combinations of items. We present two theoretical results: 1. BD in double auctions, where goods can be sold as well as bought, can be formally reduced to the problem of BD in single-sided auctions. 2. BD problems in simultaneous auctions are isomorphic to common variants of WD in combinatorial auctions. In a longer version of this paper, we propose and experiment with algorithmic approaches to BD.