On cognitive channels with an oblivion constraint

We consider a Gaussian point-to-point secondary link which co-exists with a Gaussian point-to-point primary link, and is permitted to transmit within the primary user's interference margin — a simple but realistic channel model for underlay cognitive systems. Under this assumption, we are interested in quantifying the secondary user rate gains when the secondary users have knowledge of the primary link's codebook. This may be exploited at the secondary transmitter or receiver — we focus on the latter. As first proposed by Popovski et al., this codebook knowledge may be exploited to opportunistically — when channel conditions permit — either fully cancel the primary user interference at the secondary receiver, or otherwise treat it as noise. In this work, we propose an achievable rate region in which the primary user transmits according to an information theoretic broadcast strategy, but otherwise remains oblivious to the secondary user operation. The secondary receiver exploits the primary codebooks, together with its broadcast strategy transmission to decode a portion of the primary message, thereby opportunistically partially canceling the primary user interference seen at the secondary receiver and thus straddles the two extremes between fully decoding and treating the interference as noise as presented by Popovski et al. Interestingly, it is shown that this broadcast strategy does not outperform Popovski et al.'s simple “all or nothing” strategy. We thus seek to tighten outer bounds on the secondary user's channel capacity. The inclusion of an average power constraint on the secondary link furthermore allows us to develop a new, tighter outer bound for this channel, which is numerically evaluated and compared to the inner bounds.