Successive Interference Cancellation-aided Diffusion Models for Joint Channel Estimation and Data Detection in Low Rank Channel Scenarios

Sagnik Bhattacharya, Muhammad Mohsin, Kamyar Rajabalifardi, J.m. Cioffi

Abstract

This paper proposes a novel joint channel-estimation and source-detection algorithm using successive interference cancellation (SIC)-aided generative score-based diffusion models. Prior work in this area focuses on massive MIMO scenarios, which are typically characterized by full-rank channels, and fail in low-rank channel scenarios. The proposed algorithm outperforms existing methods in joint source-channel estimation, especially in low-rank scenarios where the number of users exceeds the number of antennas at the access point (AP). The proposed score-based iterative diffusion process estimates the gradient of the prior distribution on partial channels, and recursively updates the estimated channel parts as well as the source. Extensive simulation results show that the proposed method outperforms the baseline methods in terms of normalized mean squared error (NMSE) and symbol error rate (SER) in both full-rank and low-rank channel scenarios, while having a more dominant effect in the latter, at various signal-to-noise ratios (SNR).

Introduction

In massive MIMO (Multiple-Input Multiple-Output) systems, traditional signal processing assumes a "full-rank" scenario where the base station possesses at least as many antennas as there are active users. However, as wireless communication moves toward 6G and higher frequencies (mmWave and THz), "low-rank" scenarios become prevalent. These occur when the environment is sparse or when the number of users exceeds the available antennas at the access point. In such underdetermined systems, standard methods like Minimum Mean Squared Error (MMSE) struggle. While diffusion models—generative AI techniques—have shown promise for channel estimation, they typically falter when the channel rank is low. This paper introduces a novel algorithm combining Successive Interference Cancellation (SIC) with score-based diffusion models to bridge this gap.

Methodology

The proposed SIC-aided Langevin Diffusion algorithm employs a "divide and conquer" strategy to handle rank-deficient channels:

  • Channel Splitting: The low-rank channel matrix is partitioned into smaller submatrices that are closer to full-rank, facilitating more accurate modeling by the AI.
  • Score-Based Diffusion: A generative diffusion model learns the "prior" distribution of these channel components. It uses Langevin dynamics to iteratively guide the estimation process toward realistic channel structures, filling in information gaps left by the physics of the low-rank environment.
  • Dynamic SIC Decoding: The algorithm incorporates Successive Interference Cancellation with a dynamic decoding sequence. This sequence is updated in every iteration based on current channel strength estimates, progressively simplifying the problem as stronger signals are removed.
  • Joint Iterative Optimization: The process alternates between updating channel estimates and detecting data symbols. To handle discrete QAM symbols, an annealing process creates a continuous approximation, allowing gradient-based optimization via the diffusion model.

Experimental Results

The researchers evaluated the model using the 3GPP-compliant QuaDRiGa channel model. They simulated a challenging low-rank setup with 64 users ($N_u$) and only 32 antennas ($N_r$), creating a clear rank-deficient scenario. Performance was compared against established baselines: L-DAMP (Learned Denoising-based Approximate Message Passing), LMMSE, and a standard Joint Langevin diffusion approach.

Performance comparison in high-rank scenarios showing NMSE (a) and SER (c).
Performance comparison in low-rank scenarios showing NMSE (b) and SER (d).

As shown in the results, the proposed model significantly outperforms all baselines:

  • Channel Estimation (NMSE): In low-rank scenarios (Figure b), the proposed model achieves a Normalized Mean Squared Error several decibels lower than competitors, particularly at high Signal-to-Noise Ratios (SNR).
  • Data Detection (SER): The Symbol Error Rate (Figure d) is consistently lower across the board. Even when the number of transmitted data symbols increases (from 10 to 70), the proposed method maintains a significant performance lead over the Joint Langevin baseline.
  • Versatility: While optimized for low-rank challenges, the algorithm also demonstrates superior performance in standard high-rank scenarios (Figures a and c), proving it is a robust, general-purpose solution.

Conclusion and Future Work

By leveraging learned priors through diffusion models and structural simplification via SIC, the proposed algorithm effectively handles joint channel estimation and data detection in rank-deficient MIMO systems. While highly effective at medium-to-high SNR, future research will aim to enhance performance in low SNR regimes and reduce the computational complexity inherent in iterative diffusion processes. Additionally, the authors propose integrating reinforcement learning for optimized power allocation to further boost performance in sparse environments.

Summary updated

Abstract

This paper proposes a novel joint channel-estimation and source-detection algorithm using successive interference cancellation (SIC)-aided generative score-based diffusion models. Prior work in this area focuses on massive MIMO scenarios, which are typically characterized by full-rank channels, and fail in low-rank channel scenarios. The proposed algorithm outperforms existing methods in joint source-channel estimation, especially in low-rank scenarios where the number of users exceeds the number of antennas at the access point (AP). The proposed score-based iterative diffusion process estimates the gradient of the prior distribution on partial channels, and recursively updates the estimated channel parts as well as the source. Extensive simulation results show that the proposed method outperforms the baseline methods in terms of normalized mean squared error (NMSE) and symbol error rate (SER) in both full-rank and low-rank channel scenarios, while having a more dominant effect in the latter, at various signal-to-noise ratios (SNR).
BibTeX
@inproceedings{bhattacharya2025successive,
  author = {Sagnik Bhattacharya and Muhammad Mohsin and Kamyar Rajabalifardi and J.M. Cioffi},
  title = {{Successive Interference Cancellation-aided Diffusion Models for Joint Channel Estimation and Data Detection in Low Rank Channel Scenarios}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2025},
  doi = {10.1109/icassp49660.2025.10888845},
}
Figure