Spectrum Trading for Non-Identical Channel Allocation in Cognitive Radio Networks

In this paper, we investigate spectrum trading via auction approach for exclusive usage spectrum access model in cognitive radio networks. We consider a realistic valuation function which depends on channel capacity, delay sensitive and delay-insensitive data traffics of secondary users (SUs), and propose an efficient concurrent Vickrey-Clarke-Grove (VCG) mechanism for non-identical channel allocation among r-minded bidders in two different cases. In the first case, the SUs are r-minded but they can submit bid only for single channels. In the second case, the SUs are r-minded and they can submit bid for bundles of channels. We show that the first case is solvable in polynomial time but in the other one, the problem of determining auction outcomes is NP-hard. We propose two sub-optimal methods for solving this problem, namely greedy algorithm and randomized rounding linear programming (LP) relaxation algorithm. Due to the sub-optimal nature of solutions in the second case, VCG mechanism is not truthful anymore and the SUs can lie to maximize their utilities. To address this, we propose an auction mechanism with limited truthfulness property based on an iterative greedy algorithm.