[Paper Review] On Optimal Link Activation with Interference Cancellation in Wireless Networking

Di Yuan, Vangelis Angelakis+3 more|arXiv (Cornell University)|Dec 6, 2011

Cooperative Communication and Network Coding32 references3 citations

TL;DR

This paper investigates optimal link activation in wireless networks using interference cancellation (IC), proposing integer linear programming formulations for both parallel (PIC) and successive (SIC) IC. It demonstrates that SIC can nearly double or triple the number of simultaneously active links at low to medium SINR thresholds, significantly improving spatial reuse compared to conventional single-user decoding.

ABSTRACT

A fundamental aspect in performance engineering of wireless networks is optimizing the set of links that can be concurrently activated to meet given signal-to-interference-and-noise ratio (SINR) thresholds. The solution of this combinatorial problem is the key element in scheduling and cross-layer resource management. Previous works on link activation assume single-user decoding receivers, that treat interference in the same way as noise. In this paper, we assume multiuser decoding receivers, which can cancel strongly interfering signals. As a result, in contrast to classical spatial reuse, links being close to each other are more likely to be active simultaneously. Our goal here is to deliver a comprehensive theoretical and numerical study on optimal link activation under this novel setup, in order to provide insight into the gains from adopting interference cancellation. We therefore consider the optimal problem setting of successive interference cancellation (SIC), as well as the simpler, yet instructive, case of parallel interference cancellation (PIC). We prove that both problems are NP-hard and develop compact integer linear programming formulations that enable us to approach the global optimum solutions. We provide an extensive numerical performance evaluation, indicating that for low to medium SINR thresholds the improvement is quite substantial, especially with SIC, whereas for high SINR thresholds the improvement diminishes and both schemes perform equally well.

Motivation & Objective

• To study optimal link activation under multiuser decoding with interference cancellation (IC), moving beyond traditional single-user decoding (SUD) assumptions.
• To quantify the performance gains of IC—specifically parallel (PIC) and successive (SIC) interference cancellation—over conventional SUD in concurrent link activation.
• To prove the NP-hardness of the optimal link activation problem under both PIC and SIC models.
• To develop compact integer linear programming (ILP) formulations that enable computation of global optimum solutions for IC-assisted link activation.
• To evaluate the impact of network topology, SINR thresholds, and cancellation stages on IC performance gains.

Proposed method

• Formulates the optimal link activation problem under SIC and PIC as compact integer linear programs (ILPs), enabling exact solution computation via standard solvers.
• Models the interference cancellation process such that a receiver can decode and remove interfering signals before decoding the desired signal, based on their received power and SINR thresholds.
• Introduces a multi-stage cancellation model for SIC, where up to T cancellation stages are allowed per link, with each stage removing the strongest decodable interferer.
• Uses a weighted link activation model where link weights reflect utility or data rates, and maximizes total weight under SINR and IC constraints.
• Employs a dual-based column generation approach as a benchmark for comparison, aligning with standard scheduling frameworks.
• Conducts extensive numerical evaluations on synthetic and real-world topologies (I and N datasets) with varying link densities and SINR thresholds.

Experimental results

Research questions

• RQ1 How does interference cancellation (PIC and SIC) improve the number of simultaneously active links compared to conventional SUD in wireless networks?
• RQ2 What is the computational complexity of the optimal link activation problem under SIC and PIC, and can it be solved efficiently using exact methods?
• RQ3 How do SINR thresholds and network topology (sparse vs. dense) affect the performance gains from IC?
• RQ4 What is the marginal gain of additional cancellation stages in SIC, and when does it become negligible?
• RQ5 How does the performance of PIC compare to SIC, and under what conditions is PIC nearly equivalent to SIC?

Key findings

• For low SINR thresholds (e.g., -9 dB), SIC can activate nearly all 30 links in a network, while SUD activates less than one-third, demonstrating a near-tripling of active links.
• At SINR thresholds below -6 dB, SIC activates effectively all links in dense networks, whereas SUD activates less than half, indicating substantial spatial reuse gains.
• PIC provides a relatively constant improvement of 2 to 5 additional active links across various SINR thresholds, with gains diminishing as SINR increases.
• For SINR thresholds above 3 dB, all schemes (SUD, PIC, SIC) converge in performance, as IC becomes rare due to insufficient interference power for decoding.
• The first cancellation stage in SIC yields the largest performance gain, with additional stages providing diminishing returns—especially in sparse and medium-sized networks.
• In networks with heterogeneous SINR thresholds (e.g., {-6, -3, 3} dB), SIC can nearly double network throughput, with the first stage contributing roughly half the gain.