Identifying Randomly Activated Users via Sign-Compute-Resolve on Graphs

TL;DR

This paper introduces a novel contention algorithm combining physical-layer network coding, signature coding, and successive interference cancellation to identify active users in large, randomly activated user sets without prior knowledge of their number.

Contribution

It proposes a new algorithm that improves active user detection in large networks by integrating multiple coding and cancellation techniques, with performance analysis and an activity estimator.

Findings

Algorithm effectively detects active users with high probability.

Performance analyzed in both asymptotic and non-asymptotic regimes.

Estimator accurately predicts the number of active users for tuning.

Abstract

In this paper we treat the problem of identification of a subset of active users in a set of a large number of potentially active users. The users from the subset are activated randomly, such that the access point (AP) does not know the subset or its size a priori. The active users are contending to report their activity to the AP over a multiple access channel. We devise a contention algorithm that assumes a combination of physical-layer network coding and K-out-of-N signature coding, allowing for multiple detection of up to K users at the access point. In addition, we rely on the principles of coded slotted ALOHA (CSA) and use of successive interference cancellation to enable subsequent resolution of the collisions that originally featured more than K users. The objective is to identify the subset of active users such that the target performance, e.g., probability of active user resolution and/or throughput is reached, which implies that the duration of the contention period is also not known a priori. In contrast to standard CSA approaches, in the proposed algorithm each user, active or not, has a predefined schedule of slots in which it sends its signature. We analyze the performance of the proposed algorithm both in the asymptotic and non-asymptotic settings. We also derive an estimator that, based on the observation of collision multiplicities, estimates how many users are active and thereby enables tuning of the length of the contention period.