Information Rates for Channels with Fading, Side Information and Adaptive Codewords

Area of Science:

• Information Theory
• Wireless Communications
• Signal Processing

Background:

• Generalized Mutual Information (GMI) is a key metric for evaluating achievable communication rates in fading channels.
• The availability and utilization of Channel State Information at the Transmitter (CSIT) and Receiver (CSIR) significantly impact communication system performance.
• Existing methods for GMI computation, particularly those using reverse channel models with Minimum Mean Square Error (MMSE) estimates, present optimization challenges.

Purpose of the Study:

• To compute achievable rates for fading channels using Generalized Mutual Information (GMI) under various Channel State Information at the Transmitter (CSIT) and Receiver (CSIR) conditions.
• To investigate and compare different auxiliary channel models for GMI calculation, focusing on their optimization complexity and achievable rates.
• To analyze the impact of adaptive codewords and codebook design on maximizing GMI and achieving channel capacity.

Main Methods:

• Utilized variations of auxiliary channel models with Additive White Gaussian Noise (AWGN) and circularly-symmetric complex Gaussian inputs.
• Employed reverse channel models with MMSE estimates for maximum rate computation and forward channel models with linear MMSE estimates for easier optimization.
• Applied these models to fading channels with unknown CSIT, adaptive codewords, and analyzed scalar channels with conventional codebooks modified by CSIT.

Main Results:

• Forward channel models with linear MMSE estimates offer a more tractable approach to GMI optimization compared to reverse models.
• Partitioning the channel output alphabet and using distinct auxiliary models for each partition enhances GMI and aids in determining capacity scaling.
• Developed power control policies for partial CSIR and demonstrated MMSE policy for full CSIT, validated through on-off and Rayleigh fading channel examples.

Conclusions:

• GMI provides a robust framework for analyzing achievable rates in fading channels with varying CSIT/CSIR.
• The choice of auxiliary channel model and output alphabet partitioning significantly influences GMI and capacity.
• The findings generalize to block fading channels, offering insights into capacity expressions involving mutual and directed information.