Successive integer-forcing and its sum-rate optimality

Or Ordentlich, Uri Erez, Bobak Nazer

Abstract: Successive Integer-Forcing (S-IF) is a receiver architecture that bridges the gap between low-complexity linear receivers and optimal joint decoding for Multiple-Input Multiple-Output (MIMO) channels. By combining the "compute-and-forward" philosophy with successive interference cancellation, S-IF achieves the theoretical sum-capacity of a MIMO channel. This paper outlines the S-IF framework, proves its sum-rate optimality, and demonstrates its performance through numerical examples.

1. Introduction

In wireless MIMO communications, receivers must disentangle multiple overlapping data streams. Optimal joint decoding (Maximum Likelihood) achieves the best possible rates but is computationally prohibitive as the number of antennas grows. Conversely, linear receivers like Zero-Forcing or MMSE are computationally simple but suffer significant rate loss by failing to exploit noise correlations. Standard Integer-Forcing (IF) decodes integer-linear combinations of streams in parallel, which improves performance but still falls short of capacity. Successive Integer-Forcing (S-IF) addresses these limitations by decoding combinations sequentially, utilizing noise prediction to reach theoretical limits.

2. The Successive Integer-Forcing (S-IF) Framework

The S-IF receiver operates in three primary stages:

  • Linear Filtering: An MMSE filter is applied to the incoming signal to estimate a set of integer-linear combinations ($AX$) of the transmitted signals ($X$).
  • Successive Decoding with Noise Prediction: Combinations are decoded sequentially. Once a combination is decoded, the receiver uses the estimation error to predict and subtract the noise corrupting the subsequent combination. This process is the integer-forcing equivalent of Successive Interference Cancellation (SIC).
  • Lattice Basis Reduction: To find the optimal integer matrix $A$, the authors utilize Korkin-Zolotarev (KZ) lattice basis reduction. For S-IF, the optimal $A$ is a unimodular matrix, ensuring the combinations can be easily inverted to recover the original data.

3. Theoretical Optimality and Sum-Rate Performance

The authors formalize the achievable rates in Theorems 1 and 2, proving that S-IF achieves the White-Input Mutual Information ($C_{WI}$). If the transmitter can optimize power and rate allocation, S-IF reaches the full MIMO capacity. S-IF is essentially a lattice-coded implementation of the MMSE Generalized Decision Feedback Equalizer (GDFE), a well-known theoretical benchmark.

A significant advantage of S-IF is its ability to achieve more "balanced" rate tuples compared to standard SIC, which often requires extreme rate disparities between users to reach capacity.

Achievable rate region for a 2-user Gaussian MAC comparing S-IF and standard SIC.

As shown in Figure 1, for a 2-user Gaussian MAC at 15dB SNR, S-IF significantly expands the achievable rate region (indicated by the light blue area) compared to standard SIC. It offers operating points, such as (1.44, 1.85), that are much closer to the fair symmetric capacity while remaining on the sum-capacity boundary.

4. Practical Considerations and Limitations

While S-IF is sum-rate optimal, several practical challenges remain:

  • Complexity: Finding the optimal matrix $A$ via KZ reduction is NP-hard. In high-dimensional systems, approximations like the LLL algorithm may be necessary, potentially sacrificing some optimality.
  • Channel State Information: Reaching absolute sum-capacity requires Channel State Information at the Transmitter (CSIT) for correct rate allocation.
  • Error Propagation: Like all successive schemes, early decoding errors can propagate. The use of robust, long-block-length lattice codes is assumed to mitigate this risk.

5. Conclusion

Successive Integer-Forcing provides a powerful framework for achieving MIMO capacity with manageable receiver complexity. By leveraging noise correlations through sequential decoding of integer combinations, it offers a flexible solution for high-performance wireless networks.

Summary updated

BibTeX
@inproceedings{ordentlich2013successive,
  author = {Or Ordentlich and Uri Erez and Bobak Nazer},
  title = {{Successive integer-forcing and its sum-rate optimality}},
  booktitle = {Allerton},
  year = {2013},
  doi = {10.1109/allerton.2013.6736536},
}
Figure