An Adaptive Channel Training for Massive MIMO Spatial Multiplexing Systems

One of the key techniques that have evolved into the next generation (e.g., LTE) broadband wireless standards is the multiple antenna technique. The multiple antenna technique has already become part of our daily lives, which can be clearly shown by the high LTE-smart phone possession rate in Hong Kong.Today, there is gradual movement to further increase the number of antennas to tens or potentially hundreds. These systems are often dubbed as massive multiple-input multiple-output (MIMO) systems. Promising theoretical results have been shown improved network capacity is possible. However, despite the theoretical success, many challenges have to be addressed, before any of the promised gain could be harnessed. At the very front of them are questions related to channel training overhead for large-dimensional channel estimation. Usable channel estimates at the transmitter and receiver are critical to facilitate advanced MIMO precoding techniques, such as beamforming and spatial-multiplexing. However, as the number of antennas increases substantially, the training overhead could overwhelm the channel resources allocated for data communications, resulting in significant decrease in data rates. Most massive MIMO research, however, circumvents this difficulty by focusing on the single antenna user. Hence, to provide spatial-multiplexing gain in the network, a large number of simultaneous users are always required, where the scenario does not extend to emerging multi-tiered small cell (e.g., picocell/femtocell) networks. Furthermore, restricting the receiver to single antenna precludes substantial spatial-multiplexing gain that large-scale antennas can provide. Thus, there is great interest in increasing the number of antennas both at the transmitter and receiver to leverage massive spatial-multiplexing gain. However, the achievable spatial-multiplexing gain is obviously reliant on the quality of the channel estimate.The goal of the project is to provide new approaches that greatly scale the large-dimensional channel training and channel estimation for massive MIMO spatial multiplexing systems. The approaches are constrained to only require a fixed training overhead, while still improving the quality of channel estimation. The novelty of the project lies in introducing new training signal adaptation schemes that leverage feedback information, and the optimal channel resource allocation that maximizes the system performance. The result is an upgraded channel training and channel estimation framework that greatly scales the system performance with the increased number of antennas. Such a framework can serve as a baseline for developing advanced massive MIMO systems, which could potentially revolutionize current cellular deployments by providing high throughput user access in various wireless applications.