Bid and Time Truthful Online Auctions in Dynamic Spectrum Markets

Abstract

The allocation of underutilized spectrum from primary users to secondary users in real time is likely the most promising avenue for advancing efficiency of spectrum use given the ever-increasing demand for transmission. Research in this area has focused on auctions to facilitate the distribution of spectrum, inducing truthful reporting by participants. However, most research has assumed a static or partially dynamic setting. These approaches are unable to capture that spectrum becomes available at random intervals as primary users’ needs vary across time; and, similarly, secondary users’ needs vary over time. Moreover, frequently there is flexibility regarding the time of transmission—with some transmissions being more urgent and time-sensitive than others. Therefore, existing research cannot be directly applied to such auction environments involving users with variable transmission deadlines, while preserving efficiency and truthfulness. In this paper, we present two truthful online auction mechanisms in dynamic spectrum markets that consider indefinite number of arrival of bidders with varying transmission deadlines and random availability of spectrum units over time. The first proposed mechanism SOADE assumes that the underlying distribution information of bidders and supplies is available. With that knowledge, the mechanism builds around a priority function that determines the rank of a bidder of winning spectrum at an auction considering its valuation, deadline, and uncertainty associated with dynamic arrival of bidders and spectrum availability. The second proposed mechanism xSOADE does not require any distribution knowledge. This mechanism applies bid monotonic spectrum allocation technique, determines the payment based on critical pricing, and enforces penalty rules to avoid manipulation. We prove that both the algorithms are truthful against bid and time-based cheating and individually rational through theoretical analysis and numerical simulations. Finally, we analyze the performance of these algorithms under different settings in terms of auction efficiency and auction revenue and demonstrate their effectiveness compared to prior work.