Auction-based WLAN resource allocation in game theory perspective

Abstract



Surge in WLAN development leads to urgent demand for more efficient bandwidth allocation methods. The bandwidth resource becomes more and more scarce. As a consequence, it demands both revenue maximizing and cost efficient allocation methods.

To develop this sort of allocation methods, a dilemma between the target of maximizing service quality (in terms of user revenue) and the target of cost efficient allocation must be resolved. In this thesis, two novel auction-based time-slotted WLAN bandwidth allocation models are established, namely the Combination of Single-slot Second-Price Sealed Auction (CSSA) and the Multi-slot Vickrey Auction versus Homogeneous Objects (MVAH). The two models propose a new approach to the solution of the dilemma: user revenue is formulated by not only its revenue gain or loss related to the access to the bandwidth resource, but also by a series of WLAN transmission factors, including Packet Collision Rate and Minimal Packet Transmission error rate.

In the two models, wireless bandwidth is time-slotted and bandwidth users compete for the access to these time slots. Initially, each bandwidth user joins a bidding group which is represented by its bidder. There is only one bandwidth provider in the coverage area of the WLAN. After being assigned virtual credit budget, every bidder bids for arbitrary slot combination in each bidding trial or all slots in one bidding trial, according to the bidding rule of adopted allocation model.

Non-cooperative game formulation is thereafter introduced to analyze the revenue of each bidder, revenue of the provider, and revenue of the whole allocation system. Specifically, each bidder’s revenue function is differentiated by its bid; the extreme points of this partial differential equation, i.e., the best bids for the bidder, are computed. The bidding strategy formula for each single bidder is derived from the game formulation under the bidder’s budget limit. Each bidder adjusts its bids to attain the optimal revenue obtained from the bidding strategy formula. The allocation pattern is therefore formed.

By means of calculation of the all-bidder revenue to whole system revenue ratio, the system efficiency values for both allocation models are analyzed. Conclusions can be derived from the formula of system efficiency:

1) CSSA benefits about 1 percent higher system efficiency unit than MVAH when only average throughputs of user groups differ. And both models iii-

gain optimal system efficiencies when bidder’s average throughput differs 65 times pairwise.

2) The larger the number of users connecting to each bidder, the lower the system efficiency. And CSSA model attains higher system efficiency than MVAH when user numbers of user groups differ more than 3.2 times pairwise.

In accordance with the pursuit of maximum system efficiency, new WLANs satisfying both revenue maximizing and cost efficient allocation demands can be established.