The goal of this research is to extend the theoretically important idea of a VCG (i.e. second-price, combinatorial) auction to one that is feasible in practice, by developing a mechanism that achieves the full benefits of a VCG auction but eliminates the need for bidders to bid on an exponential number of bundles and simultaneously reduces the computational complexity of the winner- and payment- determination problems. We will provide a new, mathematical programming (MP) based mechanism that captures all of the benefits of a combinatorial auction implicitly by leveraging the bidders' true cost types. In developing and investigating this mechanism, we will address issues that are both theoretical (for example, How would the approach change given stochastic input data?) and applied (such as, What applications are the most appropriate candidates for this approach? How can the MP's be solved more efficiently?). Particular attention will be paid to the real-world problem of truckload procurement.
