A Machine Learning Auxiliary Approach for the Distributed Dense RFID Readers Arrangement Algorithm

Peizhi Yan,Ruizhong Wei,Salimur Choudhury

doi:10.1109/access.2020.2977683

Peizhi Yan, Ruizhong Wei + Show 1 more

Open Access

https://doi.org/10.1109/access.2020.2977683

Copy DOI

Abstract

This paper is an extended version of the work published. Radio-frequency identification (RFID) is widespread in industries such as supply-chain management and logistics due to its low-cost feature. In many real-world problems, one often needs to leverage a considerable amount of RFID readers to cover a large area. Many graph-based dense RFID readers system anti-collision algorithms were proposed to address the collision problems. However, state-of-the-art collision avoidance algorithms are centralized algorithms. In a dense RFID system, the graphs generated by the centralized algorithms could be very complicated. Therefore, a centralized algorithm increases the computational workload of the central server. We proposed a distributed anti-collision algorithm based on the idea of a centralized collision avoidance algorithm called MWISBAII. In our later research, we found that due to the lack of global information, there is a gap between the performance of our distributed algorithm and the centralized MWISBAII. To narrow this gap, we introduced machine learning into the proposed algorithm. The machine learning model is an empirical model that mitigates the deficiency of the lack of global information. The experimental results show that the proposed distributed algorithm with machine learning can get almost the same performance as the centralized MWISBAII in different experimental settings.

Highlights

Radio-frequency identification (RFID) is an automatic identification and data capture technology, using radio frequency electromagnetic waves to transmit signals [3]
THE DISTRIBUTED MWISBAII To allow each RFID reader to be involved in the computation and decision process, we proposed a distributed MWISBAII in [1]
WORK In our previous research, we found that due to the lack of global information, there is a gap in performance between the distributed MWISBAII and the centralized MWISBAII

Summary

INTRODUCTION

RFID is an automatic identification and data capture technology, using radio frequency electromagnetic waves to transmit signals [3]. If the number of tags within an RFID reader’s interrogation range is greater than the maximum number of tags that can be read by an RFID reader (we use α to represent this limit), this RFID reader cannot be activated. This scenario can be considered as a type of collision. The problem of selectively activate or deactivate the interrogation ranges in a dense RFID readers system to allow the system to read as many tags at the same time as possible is known as reader-coverage collision avoidance (RCCA) problem [38]. The problem of this algorithm is when the number of nodes is huge, the message size could be exponentially large

GWMIN2 Algorithm

THE DISTRIBUTED MWISBAII

THE PROPOSED MACHINE LEARNING AUXILIARY APPROACH

EXPERIMENTAL RESULTS

CONCLUSION AND FUTURE WORK