Pianov D., Watts A.
11/13/2015
Radio Spectrum - the radio frequency (RF) portion of the electromagnetic spectrum. Spectrum is divided by the bands, which is allocated to the companies. In the literature it is conventional to call them primary users (PU's) of radio spectrum;
Recent legislative attempts to liberalize spectrum market and award complete property rights to the licenced primary users of the spectrum highlight a need to establish foundation of instruments and techniques to analyze the secondary spectrum sharing market (FCC 04-167, FCC 03-113, IA DCMS070);
Increasing amount of network devices pose a issue of frequency scarcity. Current dominating approach - static network allocation, has a problem of under-utilization of the bandwidth. Dynamic Spectrum Sharing (DSS) allow devices to opportunistically transmit on available radio spectrum. Some IEEE standards (IEEE 802.16, 802.22) and framework of cognitive radio support such technology;
Aside from technical difficulties of implementing DSS, it is as important to address economic part of such market interactions.
We draw from the literature in a following way:
Buyer-seller network theory (Kranton and Minehart (2001), Jackson and Watts (2002), Goyal (2009));
Secondary Spectrum Sharing (Zhang and Zhou (2014), Watts (2015), Wang et al (2010));
Auction theory (Cramton (2013))
\[ u(x_i)= \begin{cases} v_i, & \text{if $x_i=1$}\\ 0, & \text{otherwise} \end{cases} \]
Participants are randomly assigned over metric space \( \Omega \);
Every time period auction take place in the coverage areas. As a result, some of the SU's will get the good and some of them do not. We assume SU to be myopic and ultimately only caring about getting the good;
The result of the auction is price vector \( \mathcal{P}_t \);
If SU won the auction and got the good, he will remain in place for the next period. If he does not, he get the chance to reallocate to the any point of the space while incurring the transaction cost. We assume SU will move if minimal price exceed his valuation.
We propose two type of auctions:
Simultaneous Ascending Auction (SAA) - open ascending bid auction take place simultaneously in all coverage areas.
Ascending Clock Auction (ACA) - ascending auction, where buyers compete for quantities given the price.
At first, we consider only static context, restricting SU's to their initial positions.
\[ W(P, X(P,\mathcal{N})|\mathcal{N})=W_{SU}+W_{PU}, \]
with \( W_{SU} \) and \( W_{PU} \) defined as follows: \[ W_{SU}=\textbf{e}U(X(P,\mathcal{N}))-PX(P,\mathcal{N})^T\textbf{e}^T, \] \[ W_{PU}=PX(P,\mathcal{N})\textbf{e}^T-eC^T, \]
where \( X(P,\mathcal{N})=(X(P,\mathcal{L}_1)^T,...,X(P,\mathcal{L}_m)^T) \) - matrix of demand by PU, \( X(P,\mathcal{L}_j)=(x_i(P,\mathcal{L}_j)) \) - allocation vector of the good given subset \( \mathcal{L}_j \), \( U(X(P,\mathcal{N}))=(u(x_i(P,\mathcal{E})))^T \) - utility vector. Then we will call market \( \textit{statically efficient} \) at period \( t \) if
\[ P=argmax(W(P,X(P,\mathcal{N})|\mathcal{N})\text{ s.t. }\textbf{e}X(P,\mathcal{L}_j)\leqslant s_j \\ \text{ for }j=1..m\text{ and }\textbf{e}X(P,\mathcal{N})^T\textbf{e}^T\leqslant\textbf{e}S^T). \]