Channel Estimation and Feedback Techniques for Massive MIMO Systems

Abstract

To meet the demand of high throughput in next generation wireless systems, various directions for physical layer evolution are being explored. Massive multiple-input multiple-output (MIMO) systems, characterized by a large number of antennas at the transmitter, are expected to become a key enabler for spectral efficiency improvement. In massive MIMO systems, thanks to the orthogonality between different users' channels, high spectral and energy efficiency can be achieved through simple signal processing techniques. However, to get such advantages, accurate channel state information (CSI) needs to be available, and acquiring CSI in massive MIMO systems is challenging due to the increased channel dimension. In frequency division duplexing (FDD) systems, where CSI at the transmitter is achieved through downlink training and uplink feedback, the overhead for the training and feedback increases proportionally to the number of antennas, and the resource for data transmission becomes scarce in massive MIMO systems. In time division duplexing (TDD) systems, where the channel reciprocity holds and the downlink CSI can be obtained through uplink training, pilot contamination due to correlated pilots becomes a performance bottleneck when the number of antennas increases. In this dissertation, I propose efficient CSI acquisition techniques for various massive MIMO systems. First, I develop a downlink training technique for FDD massive MIMO systems, which estimates the downlink channel with small overhead. To this end, compressed sensing tools are utilized, and the training overhead can be highly reduced by exploiting the previous channel information. Next, a limited feedback scheme is developed for FDD massive MIMO systems. The proposed scheme reduces the feedback overhead using a dimension reduction technique that exploits spatial and temporal correlation of the channel. Lastly, I analyze the effect of pilot contamination, which has been regarded as a performance bottleneck in multi-cell massive MIMO systems, and propose two uplink training strategies. An iterative pilot design scheme is developed for small networks, and a scalable training framework is also proposed for networks with many cells.