In modern communication networks, measurements and signals are often fedback to various nodes across the network. This project addressed three fundamental problems in this context with respect to three important components of models, algorithm design, and fundamental limits. The work conducted at UCSD, considered the feedback in the context of sequential learning and parameter tuning in communications.  Our contributions are as follows:

1. Established an equivalence between feedback communication and sequential beamforming. 

2. Derived fundamental limits on the rate of acquiring optimal beamforming vector. 

3. Proposed optimized sequential beamforming methods inspired by posterior-matching techinque in communication systems. 

4. Characterized the gain associated with adaptive and sequential beamforming over openloop methods. 

The above findings were first obtained in very general setting applicable to a variety of application scenarios which were published in IEEE Journal of Selected Topics in Signal Processing and the Proceedings of the International Symposium on Informatin Theory. The work was also featured invited presentation at major Information Theory conferences,

We also specialized the above theoretical results in the specific context of (sub-)mmWave communications. Figure 1, for instance, shows the model of sequential beamforming for mmWave communication among UAVs. In mmWave communications, higher data rate long-range transmission is achieved by using highly directional beams with access to larger bandwidth. It is evident, however, that the quality of the communication links is directly proportional to the ability of radios to dyanamically track and refine beamforming vectors. In other words, an inherent challenge is tracking channel state information (CSI) necessary for mmWave transmission under systems with unpredictable mobility. We proposed a novel method of active and sequential beam tracking at mmWave frequencies and above. We focused on the dynamic scenario of UAV to UAV communications where the problem is equivalent to tracking an optimal beamforming vector along the line-of-sight path. In this setting, the resulting receiver beam ideally points in the direction of the angle of arrival with sufficiently high resolution. Prior work and previously proposed solutions account for predictable movements or small random movements using known filtering strategies but resort to re-estimation protocols when tracking fails due to unpredictable movements. In contrast, we proposed an algorithm for actively and sequentially selecting beamforming vectors based on a Bayesian posterior with a prediction step to account for the mobility. Numerically, we analyzed the normalized beamforming gain achieved by our proposed algorithm and demonstrate significant improvements over existing strategies. Most notably, our proposed work provided the first demonstration of a stand-alone mmWave communication system achieving sufficiently high spatial beam resoution at the  (practically relevant) low signal-to-noise ratio (SNR) regime.

The broader impact of the proposal including educting two graduate students and broadening participation among under-served communities.  

Last Modified: 03/16/2021
Modified by: Tara Javidi