On the Gaussian interference channel with half-duplex causal cognition

This paper studies the two-user Gaussian interference channel with half-duplex causal cognition, or with unilateral source cooperation. This channel model consists of two source-destination pairs sharing a common wireless channel. One of the sources, referred to as the cognitive transmitter, can gather information about the activity of the other source, referred to as the primary transmitter, through a noisy link and can therefore assist, i.e., causally cooperate, in sending the primary's data. Due to practical constraints, the cognitive transmitter is assumed to work in half-duplex mode, that is, it can not simultaneously transmit and receive. This model is more relevant for practical cognitive radio systems than the classical information theoretic cognitive channel model, where the cognitive transmitter is assumed to have non-causal a priori knowledge of the primary's message. Different network topologies are considered, corresponding to different interference scenarios: (i) the interference-symmetric scenario, where both destinations are in the coverage area of the two sources and therefore experience interference, and (ii) the interference-asymmetric scenario, where one destination does not suffer from interference. For each topology the ultimate sum-rate performance is studied by first deriving the generalized Degreesof-Freedom (gDoF), or "sum-capacity pre-log" in the high-SNR regime, and then showing relatively simple coding schemes that achieve a sum-rate upper bound to within a constant number of bits for any SNR. Finally the gDoF of the causal cognitive channel is compared to that of the classical noncooperative interference channel (to assess when cooperation is useless in practice) and to that of the non-causal cognitive channel (to identify the parameter regimes where unilateral causal cooperation attains its ideal ultimate limit).