In a combinatorial auction, multiple items are for sale simultaneously to a set of buyers. These buyers are allowed to place bids on subsets of the available items. A special kind of combinatorial auction is the so-called matrix bid auction, which was developed by Day (2004). The matrix bid auction imposes restrictions on what a bidder can bid for a subsets of the items. This paper focusses on the winner determination problem, i.e. deciding which bidders should get what items. The winner determination problem of a general combinatorial auction is NP-hard and inapproximable. We discuss the computational complexity of the winner determination problem for a special case of the matrix bid auction. We compare two mathematical programming formulations for the general matrix bid auction winner determination problem. Based on one of these formulations, we develop two branch-and-price algorithms to solve the winner determination problem. Finally, we present computational results for these algorithms and compare them with results from a branch-and-cut approach based on Day and Raghavan (2006).
|Name||Lecture Notes in Computer Science|
|Conference||WEA 2007, Rome, Italy; 2007-06-06; 2007-06-08|
|Period||6/06/07 → 8/06/07|
|Other||WEA 2007, Rome, Italy|