Design of truthful double auction for dynamic spectrum sharing

Abstract

To cope with the shortage of spectrum for wireless communications, dynamic spectrum sharing through auction mechanism is one potential approach for raising spectrum efficiency and wireless access capacity. This paper presents a design of truthful double auction for multiple heterogeneous spectrum units (TDAMH). Unlike existing double auctions for spectrum sharing, TDAMH not only supports sellers' and buyers' diverse demands in quantity but also considers the spectrum heterogeneity among sellers, which are significant trends in next generation wireless communication markets. Furthermore, TDAMH exploits the spatial reusability of spectrum by grouping conflict-free buyers and allowing buyers in one group to share same spectrum units. TDAMH has two innovations. The first is the division of buyers in one group with diverse demands into multiple virtual bidding groups (VBGs) with single unit demand, which allows each buyer to bid on multiple heterogeneous spectrum units and enables spectrum reusability. In a VBG bidding for one specific resource unit, the buyer who bids the lowest will be eliminated but the buyer's bid is selected for the determination of the VBG bid. The second is a discriminatory settlement mechanism between sellers and buyers that charges winning buyers in one VBG the bid of the VBG and pays winning sellers in one clearing price, which guarantees the properties of individual rationality and budget balance. Integration of the two innovations also incentivizes sellers and buyers to bid truthfully based on supply and demand information. Besides proofs of the three economic properties, this paper provides extensive numerical experimentation to evaluate performance of TDAMH in spectrum rent-out ratio (ROR), reusability and revenue. Results show that, by enabling spectrum reusability, although ROR is below 0.6 in all test instances, the generated spectrum revenue is higher than the expected revenue by sellers in about 40% of the total test instances.