Abstract:
Cognitive radio networks are wireless systems that exploit side information about the channel for more efficient communications. The approach of utilizing this extra knowledge is expected to lead to better use of spectrum and improved network performance. To obtain insight into this problem, we consider a building block of a network with cognitive users - a two-sender, two-receiver channel in which one "cognitive" encoder knows the other user's message. We present different regimes in which this network can operate depending on channel conditions. We show that a variety of encoding techniques - rate-splitting, Gel'fand-Pinsker coding and superposition coding becomes relevant. The obtained rates are evaluated for Gaussian channels. The numerical results demonstrate the performance gains from the proposed encoding techniques. We contrast this approach to the interweave approach in which the cognitive radio transmits in unused frequency bands. We then discuss the impact of the causal knowledge of the side information at the cognitive encoder on these techniques. Presented results offer guidelines for improving efficiency of spectrum usage in both licensed and unlicensed frequency bands.
This is a joint work with Andrea Goldsmith, Gerhard Kramer and Shlomo Shamai (Shitz).
Back to Main