ML-Aided Collision Recovery for UHF-RFID Systems

TL;DR

This paper introduces a machine learning-based collision recovery algorithm for UHF-RFID systems that significantly enhances throughput by accurately estimating tag numbers and separating signals using deep learning, outperforming traditional methods.

Contribution

The paper presents a novel ML-aided approach for collision recovery in RFID, including tag number estimation and signal separation with deep learning, improving system throughput.

Findings

High accuracy in estimating the number of colliding tags.

Deep learning-based signal separation improves decoding capability.

Throughput increases from 0.368 to 1.837 with the proposed method.

Abstract

We propose a collision recovery algorithm with the aid of machine learning (ML-aided) for passive Ultra High Frequency (UHF) Radio Frequency Identification (RFID) systems. The proposed method aims at recovering the tags under collision to improve the system performance. We first estimate the number of tags from the collided signal by utilizing machine learning tools and show that the number of colliding tags can be estimated with high accuracy. Second, we employ a simple yet effective deep learning model to find the experienced channel coefficients. The proposed method allows the reader to separate each tag's signal from the received one by applying maximum likelihood decoding. We perform simulations to illustrate that the use of deep learning is highly beneficial and demonstrate that the proposed approach boosts the throughput performance of the standard framed slotted ALOHA (FSA) protocol from 0.368 to 1.837, where the receiver is equipped with a single antenna and capable of decoding up to 4 tags.