Harnessing the Power Domain for Next Generation Interference Networks - a Bridge from Theory to Practice
Abstract
Approved for Public ReleaseObjective: Information theoretic understanding of the capacity of wireless interference networks has advanced by leaps and bounds over the past decade, spurred in large part by sharp analytical characterizations of high SNR performance, namely the Degrees of Freedom (DoF) of wireless networks with a focus on creating non-interfering channels through clever precoding across time, frequency, space and code dimensions. Despite early recognition of Han-Kobayashi schemes, a crucial aspect that has remained surprisingly underexplored in conjunction with these advances is the robust idea of optimizing interference within the power domain, through rate-splitting (RS), successive interference cancellation (SIC) combined with power control. A similar story emerges on the practical side, where after many generations of wireless networks based on frequency domain (FDMA - 1G), time domain (TDMA- 2G), code domain (CDMA - 3G), orthogonal time-frequency domain (OFDMA - 4G) and spatial domain (Massive MIMO - 5G) partitioning of resources among users, the importance of power domain partitioning through rate-splitting and successive interference cancellationalong with power control is gaining recognition, with much recent effort directed towards SIC in the form of NOMA (non-orthogonal multiple access) # albeit limited mostly to the cellular downlink and broadcast settings, while interference networks remain relatively unexplored. Recognizing its critical importance the proposed research is focused on the power-domain, as we seek to bridge the analytical insights from high SNR (signal to noise power ratio) studies with the practical algorithmic approaches at intermediate SNRs.Technical Approach: The proposal is comprised of 5 research thrusts. The first thrust explores constrained Generalized Degrees of Freedom (GDoF) characterizations, specialized to schemes that employ RS, SIC and power control along with robust interference alignment principles. The second thrust seeks the Fourier-Motzkin elimination of power control parameters to obtain closed form characterizations of GDoF regions. The third research thrust develops transformations from networks that employ SIC and RS into simpler networks that treat interference as noise (TIN) in such a manner that the two networks have the same GDoF, thus allowing access to the powerful analytical and algorithmic machinery that has previously been developed for TIN and is thus far lacking for SIC and RS. The fourth thrust navigates the curse of dimensionality inherent in the study of large networks by applying an extremal network theory approach. The final research thrust seeks representative applications of machine learning for power domain optimization in interference alignment schemes, especially at finite SNRs with discrete modulation alphabet, by exploiting insights from GDoF studies as domain knowledge for initializing the neural networks.Naval Significance: The proposed research is essential to bridge the gap between theory and practice of interference management, and to combine the benefits of power domain optimization with the dimensional optimizations of robust interference alignment schemes. As such, the research is aimed at the central idea of information dominance that is critical to future defense applications.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 09, 2024
- Source ID
- N000142412618
Entities
People
- Syed Jafar
Organizations
- Office of Naval Research
- United States Navy
- University of California, Irvine