MUMBA: Multi-unit multi-broker auctions for CRNs

Several studies have highlighted the severe under utilization of licensed spectrum by its incumbent (primary) users. To increase spectrum utilization, studies suggested using a dynamic spectrum access model in both spatial and temporal dimensions. For such a model, the spectrum is continuously monitored and the unused parts of it are allocated to secondary users. This is known as Cognitive Radio Networks (CRNs), where secondary users (nodes) send their messages (packets) directly to a nearby Secondary Base Station (SBS), which is responsible for forwarding the messages to their designated recipients. In terms of their power sources, there are two types of secondary nodes: continuous power nodes (CPN) and limited power nodes (LPN). An LPN's goal is to send its message with the minimum battery power consumption. To save power, an LPN can use a nearby “idle” CPN as a relay. On the other hand, an idle CPN (which has no packets of its own to send) aims at maximizing the benefit of its resources. These different types of nodes compete with each other to achieve their goals. An auction-based market mechanism is an appealing option to regulate such a competitive environment. It is favored due to its simplicity, efficiency and high utilization of the spectrum. The model consists of a set of brokers (idle CPNs) and an available (unused) spectrum divided into channels. The brokers iteratively issue short-term dynamic spectrum leases of these channels to competing LPNs. They choose a temporary leading broker to run the auction. The leading broker receives the LPNs' bids (which are based on their spectrum demands) and computes the output of the auction (i.e., the set of winners and their payments) with the objective of obtaining maximal revenue. Computing a solution with the maximum revenue is known to be an NP-hard problem even if there is only one broker. We design a greedy algorithm to solve this problem in polynomial time and compare it with the brute-force solution requiring exponential time. We conduct several experiments and compare the two mechanisms in terms of revenue, running time and spectrum utilization. The results show that the greedy algorithm is very fast and produces solutions that are very close (or sometimes identical) to the optimal solution.