An Amplify-and-Forward Scheme for Spectrum Sharing in Cognitive Radio Channels

Abstract

In this paper, we propose a cognitive radio scheme that allows a secondary user (SU) to transmit over the same time-frequency slot of a primary user (PU), even when the PU is active. In our scheme, the SU amplifies and forwards the signal of the PU, by using as scaling factor the value of its information symbol to be transmitted towards the secondary receiver. The information-theoretic limits of the proposed protocol are investigated in terms of ergodic channel capacities of both the PU and SU links. It is shown that: (i) under certain operating conditions, the SU can superimpose its information symbols on the PU signal, without violating the cognitive radio principle of protecting the PU transmission; (ii) when the primary link is busy, the SU offers the PU its own transmitting power in exchange for a low-capacity communication channel, which improves the packet delay performance of the SU. In this barter, the tempting incentive for the PU consists of a noticeable improvement of its achievable rate at the price of a slight increase in the computational complexity of the primary receiver.Index Terms-Amplify-and-forward relaying, cognitive radio, ergodic channel capacity, wireless communications.

Convolutive Superposition for Multicarrier Cognitive Radio Systems

“…where we have remembered that F (n) (m) = α H m x SU,I . The lower bound (28) boils down to that reported in [12], [13] when M vc = 0, L SU = 0 or, equivalently, N = 1, α m ≡ α m = 1, and γ m ≡ γ m = 0, for each m ∈ {0, 1, . .…”

“…In contrast, our scheme can achieve satisfactory data rates even without exploiting the (possible) presence of VCs, especially when the STx is sufficiently close to the SRx. In particular, when d 12 /d 14 > 0.3, the proposed scheme without VCs ensure a significant increase in data rate with respect to the NOCR scheme proposed in [12], [13] (see Fig. 7), which not only is unable to exploit the presence of the VCs, but also assumes the transmission of one SU symbol per PU data block, i.e, N = 1, and does not carry out the precoding of the SU data.…”

“…12 is a lower-triangular Toeplitz matrix having as first column (13) is fulfilled. On the other hand, it is verified by direct inspection that: (i) only the first L 12 + L SU + L 23 + θ 12 + θ 23 rows of H 23 F(n − 1) v 2 (n − 1) are located within their first L 23 + θ 23 rows.…”

“…Additionally, as a performance comparison, we report the exact ergodic capacity C PU,direct of the OCR scheme, referred to as "Conventional OCR", when the SU transmits only on the VCs of the PU, i.e., F(n) = O P ×P in (9); we also plot the worst-case capacity of the NORC scheme [13] when no CSI is available, referred to as "NOCR [13]", which is obtained from (9) by setting L SU = 0 ⇒ N = 1, G = O M ×Mvc , and A = √ a [1, . .…”

“…The channel estimation error can be made negligible, for intermediate-to-high SNRs, by using a number of training symbols that is not smaller than Lcp and by carefully designing the PU training sequence[23] 13. To suppress only its own IBI represented by uSU(n−1), the SRx could even discard a portion of the received data smaller than Lcp, thus accepting the IBI of the PU transmission due to uPU(n − 1).…”