Proceedings of Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, November 2010.
In multicell cooperation, knowledge of the users' channel state information (CSI) at the base stations can be used to improve sum-rates and reduce outages. Users estimate and feedback CSI of the desired and multiple interfering channels in cooperation-based systems. Existing multicell limited feedback techniques are memoryless quantization approaches, which result in large feedback overhead. In this paper, a novel technique is proposed using predictive vector quantization to reduce feedback requirements associated with multicell systems by exploiting temporal correlation in channels. An approach to further reduce feedback overhead by transmitting back only the normalized vector directions is also presented, using linearprediction. Simulations are used to show that the proposed predictive framework yields higher sum-rates with fewer bits as compared to existing multicell limited feedback strategies.
This preprint is available here.