Faculty Publications

Title

ReLy: Machine Learning for Ultra-Reliable, Low-Latency Messaging in Industrial Robots

Document Type

Article

Journal/Book/Conference Title

IEEE Communications Magazine

Volume

59

Issue

4

First Page

75

Last Page

81

Abstract

Robotic factory floors are transforming the manufacturing sector by delivering an unprecedented boost to productivity. However, such a paradigm raises questions on safety and coordination, especially when in the presence of unexpected events. Time-critical communication messages for such industrial robots mandate the requirement of ultra-reliable low-latency communication (URLLC). Classical WiFi-connected industrial robots often suffer from the traditional dense network problems prevalent in production WiFi networks, where transmission of an emergency notification packet is 'best effort,' devoid of time guarantees. In this work, we propose a machine-learning-based framework called ReLy that intelligently embeds the time-critical messages in the preamble of outgoing frames at the transmitter. These messages are inferred from the channel state information variations at the receiver. As ReLy is implemented entirely at the physical layer, the transmitter is able to deliver information within 5 ms latency and ultra-high reliability of 99 percent. We experimentally demonstrate the feasibility of achieving URLLC with moving robots in a busy workshop with a number of other peer robots, equipment, desks, and robotic arms, as expected in a typical factory setting.

Department

Department of Computer Science

Original Publication Date

4-1-2021

DOI of published version

10.1109/MCOM.001.2000598

Repository

UNI ScholarWorks, Rod Library, University of Northern Iowa

Language

en

Share

COinS