[Paper Review] Distributed Interference Cancellation in Multiple Access Channel with Transmitter Cooperation

I-Hsiang Wang|arXiv (Cornell University)|Sep 22, 2010

Wireless Communication Security Techniques12 references3 citations

TL;DR

This paper proposes a layered modulo-lattice transmission scheme for a two-user Gaussian multiple access channel with partial interference knowledge and transmitter cooperation, achieving capacity within 3 bits for the stronger user and 1.5 bits for the weaker user, regardless of channel parameters. The scheme enables distributed interference cancellation via cooperation, with performance robust to cooperation link capacity beyond a threshold.

ABSTRACT

We consider a two-user Gaussian multiple access channel with two independent additive white Gaussian interferences. Each interference is known to exactly one transmitter non-causally. Transmitters are allowed to cooperate through finite-capacity links. The capacity region is characterized to within 3 and 1.5 bits for the stronger user and the weaker user respectively, regardless of channel parameters. As a by-product, we characterize the capacity region of the case without cooperation to within 1 and 0.5 bits for the stronger user and the weaker user respectively. These results are based on a layered modulo-lattice transmission architecture which realizes distributed interference cancellation.

Motivation & Objective

• To characterize the capacity region of a two-user Gaussian multiple access channel with partial interference knowledge and transmitter cooperation.
• To extend prior constant-gap results to arbitrary interference powers, not just infinite interference.
• To demonstrate that cooperation provides only a bounded power gain, independent of cooperation link capacity.
• To develop a practical, distributed interference cancellation scheme using lattice-based encoding and cooperation.
• To show that the proposed scheme achieves near-optimal performance with a constant gap to capacity, regardless of channel parameters.

Proposed method

• Proposes a three-layered transmission architecture: lattice strategy layer (L), cooperation layer (C), and rate layer (R), each with distinct encoding and decoding procedures.
• Uses modulo-lattice precoding to cancel interference at the transmitter side, treating interference as structured noise in lattice-based channel models.
• Employs a cooperation layer where the stronger transmitter forwards compressed versions of its signal to the weaker transmitter via finite-capacity links.
• Applies Fourier-Motzkin elimination to combine rate constraints across layers and derive the overall achievable rate region.
• Uses Gaussian vector quantization (VQ) at the weaker transmitter to compress its signal for cooperation, though the paper notes this is suboptimal.
• Introduces effective noise terms in decoding to model interference and quantization distortion, optimizing the power allocation via parameter αC to minimize effective noise variance.

Experimental results

Research questions

• RQ1 Can a distributed interference cancellation scheme achieve a constant gap to capacity in a two-user MAC with partial interference knowledge and transmitter cooperation?
• RQ2 How does transmitter cooperation affect the capacity region when interference is known non-causally to only one transmitter each?
• RQ3 What is the impact of cooperation link capacity on performance, and does it provide unbounded gains?
• RQ4 Can lattice-based strategies outperform random binning in this interference management scenario?
• RQ5 Is the performance gap to capacity independent of channel parameters such as SNR and INR?

Key findings

• The capacity region is characterized to within 3 bits for the stronger user and 1.5 bits for the weaker user, regardless of SNR, INR, or cooperation link capacity.
• The gap remains constant even as interference powers increase, extending the constant-gap result of prior work to arbitrary interference levels.
• The cooperation link from the stronger to weaker transmitter is not required for achieving the constant gap, as it only provides a bounded power gain.
• Even with infinite cooperation capacity from Tx1 to Tx2, the capacity region increases by at most 2 bits per user compared to the non-cooperative case.
• The proposed scheme achieves near-optimal performance using layered lattice strategies, outperforming random binning-based approaches in terms of gap to optimality.
• The performance is robust to suboptimal choices such as Gaussian VQ and heuristic quantization distortion, with potential for further improvement through optimization.