Computers in Industry 112 (2019) 103127

Contents lists available at ScienceDirect

Computers in Industry
journal homepage: www.elsevier.com/locate/compind

An end-to-end Internet of Things solution for Reverse Supply Chain

Management in Industry 4.0
Celia Garrido-Hidalgo a,∗ , Teresa Olivares b , F. Javier Ramirez c , Luis Roda-Sanchez a
a
Albacete Research Institute of Informatics at the University of Castilla-La Mancha, Albacete, Spain
b
Computing Systems Department at the University of Castilla-La Mancha, Albacete, Spain
c
School of Industrial Engineering and Department of Business Administration at the University of Castilla-La Mancha, Albacete, Spain

a r t i c l e i n f o a b s t r a c t

Article history: The recent increase in the number of products returned from customers to retailers, supported by the
Received 15 March 2019 adoption of environment-friendly policies, has led to a growing need to manage backward materials and
Received in revised form 7 June 2019 information flows in the supply chain (SC) domain. Although numerous authors are contributing towards
Accepted 11 September 2019
circular economy (CE) with end-of-life (EoL) approaches minimizing the negative impact of Waste Electric
and Electronic Equipment (WEEE), the information infrastructure behind SC calls for novel approaches
Keywords:
based on Information and Communication Technologies (ICT). In fact, this is one of the major challenges
Reverse Supply Chain Management
for the so-called Industry 4.0, where wireless technologies governed by the Internet of Things (IoT)
(R-SCM)
Waste Electrical and Electronic Equipment
are expected to transform the industry as currently conceived. The present work proposes an end-to-
(WEEE) end solution for Reverse Supply Chain Management (R-SCM) based on cooperation between different
IoT IoT communication standards, enabling cloud-based inventory monitoring of WEEE through embedded
Industry 4.0 sensors. A case study was deployed using IoT devices and sensors, carrying out a set of experimental tests
BLE focused on wireless communications to evaluate its performance. The network configuration adopted
LoRaWAN overcomes the near real-time challenge and provides sufficient coverage to interconnect industrial areas
such as warehouses or shop floors. The results point to different communication bottlenecks that need
to be addressed in order to enhance the reliability of large-scale Industrial IoT (IIoT) networks.
© 2019 Elsevier B.V. All rights reserved.

1. Introduction Reverse logistics (RL) has become a key competence of both the
supply chain (SC) and reverse supply chain (RSC) [4,5], in which
The circular economy (CE) is motivated by an increasing need products flow from customers to manufacturers. Reverse Supply
to minimize the economic and environmental impact of end-of-life Chain Management (R-SCM) emerges as a new challenge, given the
(EoL) products [1]. This industrial strategy refers to the long-lasting necessity of managing every single stage where products should
design, refurbishment, remanufacture, repair, recycling or reuse be handled and distributed towards manufacturers, with multiple
of products in order to achieve the maximum benefit and avoid recovery options increasing the uncertainties faced by companies
negative impacts, with Waste Electrical and Electronic Equipment [6].
(WEEE) being a growing concern [2]. According to the United The introduction of Information and Communication Tech-
Nations University (UNU), 44.7 million metric tonnes of e-waste nologies (ICT) in SC and, especially, the Internet of Things (IoT)
was generated in 2016, with future estimations being even larger can significantly improve process-oriented performance, reduce
[3]. While posing significant risks for the environmental and human energy consumption and provide SC with a ubiquitous informa-
health, it contains recoverable raw materials with an estimated tion infrastructure [7]. By enabling the cooperation of wireless
value of 55 billion euros. technologies, IoT is an accelerator of the Fourth Industrial Revolu-
tion: Industry 4.0 [8,9]. This brand-new concept is transforming the
industrial paradigm, being Cyber-Physical Systems (CPS), cloud and
fog computing essential pillars [10,11]. The impact of IoT on indus-
∗ Corresponding author. Address: Albacete Research Institute of Informatics at tries is such that the concept of Industrial IoT (IIoT) is becoming
Investigacion Street, 02071 Albacete, Spain. increasingly popular and has consequently been adopted in several
E-mail addresses: [email protected] (C. Garrido-Hidalgo), industrial applications governed by IoT technologies [12].
[email protected] (T. Olivares), [email protected] (F.J. Ramirez),
[email protected] (L. Roda-Sanchez).

https://doi.org/10.1016/j.compind.2019.103127
0166-3615/© 2019 Elsevier B.V. All rights reserved.
2 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127

2. Related works
Nomenclature
This section provides a review of the most relevant works and
AWS Amazon Web Services applications of computer-based systems in the industrial sector,
BW Bandwidth highlighting the main contributions and gaps that encourage us to
BLE Bluetooth Low Energy propose and evaluate an end-to-end heterogeneous IoT framework
CE Circular economy for R-SCM purposes.
CR Coding rate
CPS Cyber Physical Systems 2.1. Circular economy
DR Data rate
DSC Digital supply chain The principles of CE were first outlined by Kneese et al. [16]
EoL End of life dealing with the management of residuals and the improvement
ETSI European Telecommunications Standards Institute of environmental quality. Recently, its scope has been broadened
IIoT Industrial Internet of Things to not only involve the minimization of natural resources but also
ICT Information and Communication Technology the ability to reuse materials [17]. Currently, resources are mined,
IT Information Technology turned into products and finally discarded [18]. As a result, our soci-
IoT Internet of Things ety is generating scarcity, volatility, unaffordable manufacturing
JSON JavaScript Object Notation costs and, what is more worrying, an unprecedented absence of
LoRaWAN Long-range wide-area network natural resources [19]. According to Vella [20], a shift towards a cir-
LPWA Low-Power Wide-Area cular economy could bring savings of D 600bn for EU businesses, and
MQTT Message Queing Telemetry Transport reduce greenhouse gas emissions by 2–4% every year, which highlights
RFID Radio-Frequency IDentification the necessity of adopting CE strategies.
RSSI Received signal strength indicator In this context, WEEE is receiving special attention, being one of
RL Reverse logistics the fastest-growing waste streams with a 33% increase in volume
RSC Reverse supply chain reported during the last five years [21]. The roots of this increase are
R-SCM Reverse Supply Chain Management connected with exponentially-growing consumer demand, tech-
SNR Signal-to-noise ratio nology uptake, and shorter replacement cycles. Moreover, the
SBC Single-board computer adoption of free-shipping policies on part of most companies has
SF Spreading factor accelerated electronic commerce, where the product-return ratio
STD Standard deviation has risen to 30%, more than three times that of brick and mortar
SC Supply chain sales [22].
SCM Supply chain management In the European Union, e-waste is regulated by the Directive
TTN The Things Network 2012/19/EU of the European Parliament and Council [23]. Accord-
UNU United Nations University ing to a recent communication, in 2015 only four European Union
WEEE Waste Electric and Electronic Equipment Member States recycled over half of the electronic products put
on the market [24]. This is supported by Eurostat [25], which pub-
lished a report concluding that the lowest recycling rate of waste
streams corresponds to e-waste (32%) followed by plastics (40%). In
Despite the large body of literature addressing IoT applications fact, 44.7 million metric tonnes of e-waste was generated globally
in the supply chain management (SCM) domain, the RSC field in 2016, with an annual growth rate up to 5% on a weight basis [3].
specifically needs more integrated approaches based on IoT com- Apart from the need to mitigate environmental pollution, the eco-
munication standards to manage the RL of WEEE in real contexts. nomic value associated with raw materials present in such e-waste
To fill this gap in the literature, this paper presents a novel Industry is estimated at 55 billion euros [3].
4.0 end-to-end IoT framework as a solution for WEEE management. In pursuit of a reduction of waste streams through an
The following aspects of this work can be highlighted. First, the environment-friendly and cost-saving approach, remanufacturing
proposal of a heterogeneous IoT network enabling low-power and has become one of the most feasible EoL strategies to restore
low-cost SCM operations in the context of Industry 4.0. For this, we products to ‘like-new’ condition. If well-organized and planned,
propose the cooperation of three emerging IoT technologies for R- it can expand the life-cycle of products and increase profits of
SCM, presenting a case study based on the recovery of WEEE from manufacturing companies through reducing landfills [26]. Despite
computer-based components. Second, the implementation of an being under-recognized in the current industrial landscape, reman-
end-to-end system, addressing the deployment of sensor-nodes, ufacturing generated 29.8 billion euros in turnover and 192,000
the network infrastructure, and its integration with a cloud-based employment in the European Union in the year 2015, with expec-
inventory-management platform. In order to evaluate end-to-end tations of attaining an annual value of 100 billion euros in 2030
performance, a set of experimental tests are proposed, carried out [27].
and discussed. As part of the European Research Framework H2020, the Euro-
The three wireless technologies selected and addressed in this pean Remanufacturing Network has facilitated the development
work are RFID [13], Bluetooth Low Energy (BLE) [14] and LoRaWAN of sustainable partnerships between organizations, similarly to
[15], which are tested under different communication schemes and the ResCoM (Resource Conservative Manufacturing) initiative and
physical-layer parameters. Our results illustrate the suitability of its successor ReCiPS (Resource-Efficient Circular Product-Service-
the proposed IoT standards for R-SCM purposes, underlining com- Systems).
munication optimizations for large-scale industrial deployments. In this context, information-sharing capabilities enable coop-
The paper is organized as follows: in Section 2 we review the eration among suppliers, manufacturers and recyclers, being
literature; the IoT framework for R-SCM is presented in Section 3; necessary to take responsibility for the collection, disposal and
Section 4 presents a WEEE-focused case study; Section 5 addresses recovery of products at their EoL [28]. In the literature, Garcia-
the experiments, results and discussion; and finally, Section 6 pro- Muiña et al. [29] explore the transition stages towards circularity
vides conclusions and identifies future research lines. and propose a theoretical ICT-based framework where materi-
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 3

als are sourced to manufacturers following a CE model. A similar tries in terms of Industry 4.0 adoption, which is a potential barrier
approach is provided by Asif et al. [30], who design an ICT frame- to advancement [57]. Two major areas of research are highlighted
work for information-sharing in pursue of a product-as-a-service based on the gaps that our work aims to fill in the Industry 4.0 field:
model in the washing-machine industry. Finally, Gu et al. [31] pro- heterogeneous approaches based on wireless technologies collab-
vide a theoretical IoT framework for WEEE management but lacking oration and end-to-end applications oriented towards specific use
ICT details and practical implementations. cases.
None of these works focus on using IoT technologies to address As heterogeneous approaches, Gioia et al. [58] present an IoT
information flows being exchanged upwards the SC or evaluate the gateway integrating multiple sensors, buses and technologies,
performance of communications, which we believe is a necessary while Garrido-Hidlago et al. [59] describe the implementation of a
condition. Our work, additionally, focuses on WEEE and provides hybrid network to improve working conditions in industry through
a detailed framework describing the handling stages of products the use of industrial wearables [60]. Regarding end-to-end appli-
from their EoL stage to their reuse in a R-SCM context. cations, Tao and Qi [61] propose an IT-driven framework towards
smart manufacturing. In the domain of SCM, Abdel-Basset et al.
2.2. Digital supply chain [62] present a framework for IoT-enabled securely-efficient sys-
tems, Papert and Pflaum [63] define an ecosystem to encourage IoT
The SCM strategic objectives have been widely defined in the lit- services in the domain of SCM and Wakenshaw [64] presents an IoT
erature as cost, quality, speed, flexibility and dependability [32–34]. integration framework.
During recent years, sustainability has gained special interest, with Table 1 provides a comparison of the most representative works
insufficient communication capabilities being a major obstacle to in our literature review. Many of these studies address descriptive
its adoption [35]. In this context, a number of SC information- approaches or simulations, whereas others focus on the experi-
infrastructure requirements have emerged, namely: flexibility, cost, mental evaluation of a single IoT standard lacking of an end-to-end
time, quality, accuracy, reliability, visibility and availability [36,37]. perspective or industry-oriented deployment concerning a hetero-
The digital supply chain (DSC) is thus a growing area of research geneous IoT ecosystem. We fill this gap in the literature proposing
interest, calling for innovative approaches based on emergent and testing a real end-to-end IoT solution based on a heterogeneous
technologies to manage SC processes. Its major challenges were network proposal for R-SCM of WEEE.
recently defined by Büyüközkan and Göçer [38], converging in:
speed (ability to react quickly to demand), flexibility (agile reac-
3. IoT framework for R-SCM
tion), global connectivity (internet-enabled SC), real-time inventory
(continuous monitoring of stocks levels), intelligence (self-learning
This section describes an IoT R-SCM framework proposal to
smart products), cost-effectiveness (use of technology to increase
manage WEEE, aimed at providing manufacturers with a novel
organizational performance), transparency (adjusting networks to
ubiquitous information infrastructure behind SC for tracking of
changing scenarios), scalability (optimization and duplication of
parts to be recovered. For this, we first select the set of IoT standards
processes), innovation (in pursue of competitiveness and excel-
for the deployment (based on a communication-range criterion)
lence), proactivity (anticipating issues prior to occurrence) and
and, then, the main stages of the framework are described accord-
eco-friendly (environmental emphasis).
ing to their functionality and information flows.
Information and material flows need to be managed efficiently
Our proposed IoT-based WEEE management framework is sup-
in downstream and upstream directions [18,39,2]. The main differ-
ported by a twofold approach: on the one hand, the presentation
ences between forward and reverse SCs lie in distribution channels
of an industrial solution to contribute to filling a real gap of WEEE
(standardized vs. exception-driven), transparency (asset-tracking
management across SC and, on the other, the provision of an IoT
information available vs. poor asset visibility) and timing require-
network architecture to connect SC with a transparent inventory-
ments (time-critical vs. non-time-critical), among others [40].
management infrastructure in the context of Industry 4.0.
2.3. The role of IoT in the Fourth Industrial Revolution
3.1. Selection of IoT standards
IoT, a term first coined by Ashton [41] with regard to SCM, pro-
vides physical objects – or Things – with sensing and ubiquitous The proposed R-SCM framework distinguishes between three
communication capabilities. Despite being initially conceived for sets of IoT standards according to their communication range:
customer-oriented applications, the latest advances have enabled short-range, intermediate-range and long-range. MQTT [66] was
a broader scope for IoT, permitting the integration of sensors in selected for data transmission between physical devices (IoT net-
industrial facilities, equipment and resources towards the so-called work deployed in the facilities) and virtual data infrastructures
IIoT [42,43]. In this context, the Fourth Industrial Revolution – or (inventory cloud platform), while a unique standard was selected
Industry 4.0 [44] – emerged incorporating CPS into traditional fac- from each set to fulfill the framework requirements:
tories governed by IIoT, which promises great advances [45,46]. Short-range communication. We selected high-frequency RFID
Several authors provide IoT approaches to wireless com- [13] over NFC [67] for the identification of returned products in the
munications [47]. The BLE standard has been widely studied framework. Despite being popular for contactless payment, infor-
from energy-efficiency [48,49] and low-latency [50] perspectives. mation sharing and user-oriented applications, NFC is not suitable
Rondón et al. [51] provide an in-depth evaluation of BLE, supporting for industrial use. High and ultra-high frequency RFID offer robust-
its suitability for time-critical industrial applications. Likewise, the ness against metal interference and a greater communication range
performance of the LoRaWAN standard was studied by Sanchez- (typically in the order of tens of meters).
Iborra et al. [52]; important studies also exist on improving capacity Intermediate-range communication. BLE [14] was selected over
[53] and communication performance [54,55]. Wi-Fi [68] or ZigBee [69] for local data transmission and inventory
Hofmann and Rüsch [56] regard real-time information flows, end- management in the framework. Although operating in the same
to-end transparency and flexibility as three key objectives of Industry frequency bands, BLE is oriented toward IoT applications where
4.0, whereas Büyüközkan and Göçer [38] highlight the necessity the volume of information to be sent is small and low power is
of integrating sensor networks in heterogeneous environments encouraged, having become a disruptive standard for Industrial IoT
towards this goal. Nevertheless, there is great disparity across coun- applications.
 4
C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127
Table 1
Comparison of the main contributions found in the literature.
Source Method Scope or domain area End-to-end ICT infrastructure Heterogeneous IoT network

Descriptive Theoretical Case study Simulation Implementation Experimental Industry 4.0 Internet of Circular Reverse WEEE Framework Perception Network Application Performance RFID- BLE- LoRaWAN- IP-based
Things economy logistics layer layer layer test based based based
√ √ √ √ √ √
Abdel-Basset et al.
[62]
√ √ √ √ √
Asif et al. [30]
√ √ √ √ √ √ √
Cattani et al. [54]
√ √ √ √
Cho et al. [50]
√ √ √ √
Cuomo et al. [55]
√ √ √ √
Dementyev et al.
[49]
√ √ √ √ √ √
Garcia-Muiña et al.
[29]
√ √ √ √ √ √ √ √ √ √
Garrido-Hidalgo
et al. [59]
√ √ √
Genovese et al. [17]
√ √ √ √ √
Gioia et al. [58]
√ √ √ √ √ √ √
Gu et al. [31]
√ √ √ √
Papert and Pflaum
[63]
√ √ √ √ √ √ √
Roda-Sanchez et al.
[60]
√ √ √ √ √
Rondón et al. [51]
√ √ √
Sanchez-Iborra
et al. [52]
√ √ √ √ √ √
Siekkinen et al. [48]
√ √ √ √
Tao et al. [65]
√ √ √ √
Wakenshaw [64]
√ √ √ √ √ √ √ √
Yang et al. [28]
√ √ √ √
Zorbas et al. [53]
√ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √ √
Our work
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 5

Fig. 1. R-SCM IoT framework based on: Identification Stage (i), Classification Stage (ii), Smart Containers (iii), Hybrid Gateway (iv), LoRaWAN Gateway (v), TTN (vi), AWS
(vii).

Long-range communication. Regarding Low-Power Wide-Area Table 2

IoT standards used (classified as input or output).
(LPWA) standards, LoRaWAN [70] was selected over NB-IoT [71]
or Sigfox [72] for environmental monitoring of assets and inter- Framework stage Input standards Output standards
connection of industrial facilities. The main advantages of NB-IoT (i) Identification Stage RFID BLE 4.0
and Sigfox technologies lie in their narrowband implementation (ii) Classification Stage BLE 4.2 None
providing robustness against noise and, specifically, Sigfox has the (iii) Smart Containers None BLE 4.0
lowest chip cost and NB-IoT the lowest network latencies. However, (iv) Hybrid Gateway BLE 4.2 LoRaWAN, MQTT
(v) LoRaWAN Gateway LoRaWAN IP
LoRaWAN is the best choice for bidirectional communication and
(vi) TTN IP None
provides more flexibility, enabling firmware upgrades over-the-air, (vii) AWS IoT Core MQTT MQTT
a unique feature of this technology oriented toward large-scale IoT
deployments. Furthermore, LoRaWAN provides better robustness
against mobility, which we regard as an essential requirement for fied into separate containers, the Hybrid Gateway is responsible
SC scenarios. for collecting data and managing two flows of information: global
(inventory updates) and local (environmental information).
Table 2 summarizes the compliance with IoT standards: the
3.2. Description of the stages Identification Stage (i) combines RFID technology to scan tagged
products with BLE to forward the information to the Classifica-
The proposed framework offers end-to-end IoT integration com- tion Stage (ii), received via BLE 4.2. The Smart Containers (iii) use
bining physical devices and sensors, wireless communications BLE to update inventory and environmental information, which is
(RFID, BLE and LoRaWAN) and platforms for data storage and ana- received by the Hybrid Gateway (iv) and respectively forwarded
lytics. Fig. 1 shows the framework, the main components of which through the LoRaWAN Gateway (v) to The Things Network [73],
are the Identification Stage, Inspection Stage, Classification Stage, TTN (vi) or via MQTT to IoT Core platform in Amazon Web Services
Smart Containers and Hybrid Gateway. [74], AWS (vii).
This work is aimed at deploying the communications infras- Identification Stage (i). The aim of this stage is to read informa-
tructure able to manage data gathered during the recovery of tion written on RFID tags to identify the reason of return and the
WEEE, with the Inspection Stage being outside the scope of the other key aspects of products. Tags are expected to be attached
concerned network. Nevertheless, this stage is responsible for the by the retailers or manufacturers shipping the EoL products, and
disassembly of products and the refurbishment of modular compo- written according to an agreed data-encryption algorithm – as,
nents to ensure that they meet the necessary conditions for reuse. for example, the one proposed by ElMahgoub [75] – in order to
Non-recoverable components are sent to repair or recycling plants control data access and simplify the identification of products
depending on their deterioration. Once reusable parts are classi- entering the system. Once read, an actuator system would intro-
6 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127

Fig. 3. Disassembly stages: Identification (i), Classification (ii) and Smart Containers
(iii).

• Global information flow. This consists of inventory data updated

by Smart Containers. The Hybrid Gateway, acting as a BLE scanner,
receives a mixed information flow, filters it and publishes inven-
tory updates via MQTT in AWS IoT Core (vii). Once data is stored
in AWS, any authorized entity could subscribe to get immediate
inventory updates.

4. Case study
Fig. 2. Algorithm used for classification and communication.

Personal computers have become one of the major concerns

duce reusable products into containers and let defective products regarding waste-streams generation, given a considerable decrease
continue through the system. At the Identification Stage the device in their average useful life over the years and a high environmental
is provided with RFID and BLE modules. Once the RFID tags are impact associated with their disposal. Nevertheless, most com-
scanned, this device (under the slave BLE role) exchanges informa- ponents found in computers are in good condition for reuse or
tion with its master (Classification Stage). refurbishment and, depending on their added value, these can be
Classification Stage (ii). This stage is the core of the framework, brought back to the required degree of quality as presented in [26].
where information about products and parts is collected. The parts Following this idea, the scenario selected for the case study was
received at this stage will have passed a quality control test, and are an industrial environment focused on the recovery of computer-
thus assumed to be reusable. The device located in this stage acts based products. Fig. 3 represents the flow of products and parts
as BLE master and receives information from the previous stage by throughout the stages of the system in our case study scenario.
discovering the services and reading the characteristics of the slave The two subsets of products used in the experiments were
(according to the BLE specification in Bluetooth SIG [14]). The Clas- desktop and laptop computers, which entered the system being
sification Stage includes a database where standard part models identified via RFID. Then, products would be sent to the Inspec-
are stored to access information such as reference models, kind of tion Stage, where the appropriate EoL strategy would be applied
product or weight. Subsequently, the device weighs parts to clas- according to their added value and condition (outside the scope
sify them. Fig. 2 provides a detailed description of the classification of the case study). For classification, we used components disas-
algorithm, where producttype refers to a product subset (kind of sembled from damaged computers, which are shown in Table 3, in
product, i.e. laptop or desktop computer) and parttype to the spe- addition to their approximate cost and main specifications used by
cific models of the parts stored with additional information in a the classification algorithm (see Fig. 2).
JSON file. The case study consisted in the identification and classification
Smart Containers (iii). These devices integrate a microcontroller of components using sensors and communication modules, which
platform, a load cell and an accelerometer, being able to commu- were attached to the microcontroller-based platforms deployed
nicate via BLE under the BLE advertiser role to update inventory throughout the R-SCM system. The implementation and deploy-
and context information. This stage is designed to weigh parts ment of a real hardware prototype permitted the evaluation of IoT
and calculate the available number of parts of each type to be communications taking place in the end-to-end system.
reused, turning containers into useful real-time platforms. Every The single-board computers (SBC), sensors and radio communi-
time acceleration suffers a significant change, an interruption event cation modules used to deploy the prototype are shown in Table 4.
wakes up the microcontroller and measures weight: if it has LightBlue Bean as an implementation of the ATmega 328P micro-
changed, it will recalculate the number of items (according to the controller with the LBM313 BLE radio was used for Identification
type of parts inside each container) and send a packet via BLE (see Stage (i) and Smart Containers (iv); LoPy as an implementation of
Fig. 2). Xtensa LX6 including the BLE/Wi-Fi ESP32 and the LoRa SX1276
Hybrid Gateway (iv). This device is intended to collect informa- radio module was used to develop the Hybrid Gateway; a Raspberry
tion emitted by BLE advertisers, the Smart Containers, (for which Pi 2B based on the BCM2836 microprocessor was used to deploy the
it assumes the role of scanner) and classify it according to a LoRaWAN gateway (on the roof of the testing building); a Raspberry
local/global criterion. Two kinds of information flows are created Pi Zero W based on the BCM2835 and the CYW43438 BLE radio was
in the Hybrid Gateway (see Fig. 1): used for the Classification Stage. Additionally, we attached two type
of sensors to LightBlue Bean: load cells and ambient sensors, specifi-
• Local information flow. This information is generated in Smart cally, three two-kilogram straight load cells and Pysense to test the
Containers when something unexpected occurs (i.e. low battery proposal using the temperature-and-humidity integrated sensor
level of devices) and is sent to the Hybrid Gateway. This device Si7006-A20. Regarding radio modules, the RC522 RFID communi-
is responsible for encapsulating this context data, adding sensed cation module was included at the Identification Stage, operating at
ambient information, and sending it through a LoRaWAN gate- high-frequency (13.56 MHz) and the iC880A LoRaWAN concentra-
way (v) to TTN (vi). tor (integrating the SX1301 transceiver) with an external 868 MHz
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 7

Table 3
Specifications of WEEE parts used in the experiments.

WEEE Part type Part model Subset Weight (g) Cost (D )

Hard disk ST31000340AS Desktop 642 90

Network card HE-012D Desktop 60 10

Network card FT16-04B3 Desktop 55 15

Network card GTS FC-515 Desktop 56 9

Graphic card QUADRO FX-580 Desktop 220 139

RAM KTM2865-SR Desktop 18 17

RAM CF81037 Laptop 8 40

Table 4 and scaled according to real-world experiments with IoT devices

Hardware used in the deployment (radio specifications are marked as input(i) or
instead of simulations.
output(o) information flows).
The TNSC metric quantifies the influence of increasing Smart
Framework stage Core Sensors Radio modules Containers on latency (BLE standard); TOA the Time-on-Air at
(i) Identification Stage ATmega 328P None MFRC522(i) , LBM313(o) LoRaWAN end-devices as a function of payload, complying with
(ii) Classification Stage BCM2835 TAL 220 CYW43438(i/o) ETSI [77] regulations of 1% duty cycle, NPCK ; T1 , T2 , T3 provide a
(iii) Smart Containers ATmega 328P TAL 220 LBM313(o) disaggregated end-to-end communication latency considering the
(iv) Hybrid Gateway Xtensa LX6 Si7006-A20 ESP32(i/o) , SX1276(o)
cooperation of all technologies; RSSI (received signal strength indi-
(v) LoRaWAN Gateway BCM2836 None SX1301(i) , Ethernet(o)
cator) and SNR (signal-to-noise ratio) determine the signal quality
perceived at the LoRaWAN Gateway depending on end-device loca-
tion.
Table 6 showcases the setup of the experiments, where Ga is
the antenna gain, Ptx the transmission power and Srx the receiver
sensitivity. The conditions for the Geographic Spread experiments
were cloudless skies and temperatures from 26 to 30 ◦ C, in Albacete,
Spain. The height of the transmitting end-device was 2 meters,
and the height of the LoRaWAN gateway 18 m. The LoRaWAN
specification [15] provides further information on physical-layer
parameters: spreading factor (SF), coding rate (CR), data rate (DR)
and bandwidth (BW); and the BLE specification [14] provides more
details on the main network roles used in this work, namely: scan-
Fig. 4. Prototype developed for evaluation of the R-SCM framework, showing: Iden-
tification Stage (i), Classification Stage (ii), Smart Containers (iii) and Hybrid Gateway ner/advertiser in broadcasting mode and master/slave in connection
(iv). mode.

5.1. Data ingest test

antenna was used for the LoRaWAN gateway deployment (located
outdoors, on the roof of the Albacete Research Institute of Infor-
A scaled system would imply the addition of new lines with
matics).
incoming products and, thus, the increase of BLE traffic received by
All the materials (except the LoRaWAN gateway) were dis-
the Hybrid Gateway and forwarded via MQTT to AWS. Whereas
tributed and screwed to a wooden board in order to implement
the former could become a potential bottleneck for the sys-
a low-cost prototype that permitted us to test the communication
tem, the latter has been widely tested in the literature with no
performance of the different standards. Fig. 4 shows the resulting
apparent scalability limitations for IoT, reaching up to 263,314
prototype.
packets received and 23,184 sent per hour under a 0% packet loss
[78].
5. Results and discussion A series of BLE-based experiments was carried out to evalu-
ate the influence of increasing the number of Smart Containers,
The experiments were designed to satisfy a set of performance updating inventory information to a same Hybrid Gateway. For
tests to assess the strengths and limitations of the R-SCM pro- this, the NRF Connect [79] software was used to ingest BLE traffic
posal under IoT criteria [76]. These tests are described in Table 5 in the network and measure the resulting TNSC . Each experi-
8 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127

Table 5
IoT-oriented tests to evaluate the prototype.

Stages involved Test Description Metrics

(iii), (iv) Data ingest Influence of traffic on performance TNSC

(iv), (v) Data size Influence of payload on performance TOA , NPCK
(i) → (iv), (vii) Network latency Disaggregated end-to-end performance T1 , T2 , T3
(iv), (v) Geographic spread Influence of distance on signal quality RSSI, SNR

Table 6
Experiments setup: parameters and IoT standards specifications in the R-SCM solution.

Parameter RFID Bluetooth low energy (BLE) LoRaWAN

Role Reader Master/Slave Advertiser Scanner End-device Gateway

Frequency 13.56 MHz 2.4 GHz 2.4 GHz 2.4 GHz 868 MHz 868 MHz
Modulation ASK GFSK GFSK PSK LoRa LoRa
Radiation Directional Directional Omnidirectional Omnidirectional Omnidirectional Omnidirectional
Ga (dBi) – – 1.7 – 3 8
Ptx (dBm) Variable 8.5 3 7.5 14 14
Srx (dBm) – −97 −93 −98 – −136/−120

ing time divided by the advertising interval. Eqs. (1) to (3) provide
the tendency lines and curves obtained from the experiment data:


TNSC (100ms) = 0.0129 × NSC + 0.0183 R2 = 0.9547
(1)
−0.530
T̂NSC (100ms) = 1.0323 × NSC R2 = 0.9687


TNSC (250ms) = 0.0189 × NSC + 0.4669 R2 = 0.8551
(2)
−0.416
T̂NSC (250ms) = 0.9663 × NSC R2 = 0.9206


TNSC (1000ms) = 0.0529 × NSC + 1.8576 R2 = 0.8326
(3)
−0.368
T̂NSC (1000ms) = 0.9766 × NSC R2 = 0.9285

The advertising interval implies the reception of one packet

of each container every 100 ms for a low-latency configuration,
250 ms for balanced and 1000 ms for low power. However, as the
number of Smart Containers increases, we observe in the results
how the proportional time required to receive a single packet also
increases (see Fig. 5a). Let us consider the low-power configuration
as an example: with devices advertising every 1000 ms, the ideal
TNSC would be 1000 ms, but experimentally ranges from 2000 to
4500 depending on NSC .
If different advertising intervals are considered for the same
NSC , taking into account Fig. 5b, important conclusions are reached.
Let us consider the worst-case scenario with 50 containers trans-
mitting simultaneously (NSC = 50): for a low-latency configuration,
the ideal TNSC , 100 ms, is up to 10 times lower than the experi-
mental one (nearly 1000 ms according to Fig. 5a), whereas for a
low-power configuration this value is approximately 5 times lower
Fig. 5. Influence of the increase of Smart Containers NSC in the minimum receiving
than the expected (1000 ms ideally with respect to the nearly 5000
time (TNSC ) for BLE advertising intervals of 100, 250 and 1000 ms. ms reached experimentally). This can be observed in Fig. 5b, where
the normalized minimum receiving time (T̂NSC ) for a low-power
configuration is two times that of low-latency.
Thus, despite the low-power configuration having a worse abso-
lute latency, the normalized results show that in this case the
ment was carried out ten times increasing traffic conditions under effectiveness of transmissions is the highest. This means that the
low-power (1000-ms advertising), balanced (250 ms) and low- proportion of packets lost during communication increases with
latency (100 ms) configurations. The resulting average is shown in the number of containers (NSC ) and also with the reduction of the
Fig. 5a. advertising interval. The origin of the problem may be related to col-
The results provided in Fig. 5a highlight the communication lisions taking place when the BLE traffic in the network is increased
time required in case of simultaneous transmission attempts on (either augmenting the transmitting frequency or the number of
the part of Smart Containers. Fig. 5b provides the normalized min- devices in the network). Further research is required to assess the
imum receiving time of the Hybrid Gateway for a given number of origin of this packet loss in pursuit of more efficient large-scale
Smart Containers, obtained as the inverse of the minimum receiv- industrial networks.
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 9

Fig. 6. Influence of LoRaWAN end-devices packet size (CR of 4/5, BW of 125 kHz and different SF) in TOA and NPCK for a 1% duty cycle.

5.2. Data size test significant probability of failed transmissions due to intra-SF colli-
sions (collisions occurring when two devices transmit at the same
This test focuses on the LoRaWAN network, since the per- time, using the same frequency band and SF). Additionally, inter-
formance of LPWA is significantly influenced by variable data collisions can also be noticed when the same frequency band is
payloads. In contrast to the data ingest test, where scalability shared but different SFs are set up, although this is less likely to
implied a scenario with increased traffic, this experiment aimed occur [53]. In view of the TOA results obtained, high SF configura-
to highlight the required Time-on-Air (TOA metric in Table 5) of tions and long payloads should be avoided, when possible, in order
LoRaWAN end-devices for different configurations and volumes of to reduce the probability of collisions caused by a long channel-
data. occupancy time.
Following the standard [15], the LoRa physical layer uses several The results are useful to determine the maximum number of
chips to encode each symbol, resulting in SF configurations ranging LoRa-based transmissions permitted for any end-device according
typically from SF7 (27 chips/symbol) to SF12 (212 chips/symbol). In to 1% duty cycle regulations (see ETSI [77] for further documenta-
order to calculate the Time-on-Air required when setting different tion). Fig. 6b provides a useful approximation, based on the results
SFs, the symbol duration (Tsym ) should be defined as shown in Eq. from Fig. 6a. Let us consider a 20-byte payload configured under
(4): SF10 as an example: in this case, ETSI would limit an end-device to
transmitting, at most, 2000 packets per day (more than sufficient
2SF for environmental monitoring in distributed warehouses). How-
Tsym = (4)
BW ever, TTN restrictions of 30 s of TOA per device and day [81] applied
Eq. (5) provides the formula to obtain the TOA metric (see [80] to Fig. 6a results would only permit 60 packets in the previous
design guide). The payloadSymbNb was computed considering the situation. For this reason, further research is required to use alter-
presence of a header. The preamble duration was obtained accord- native self-deployed network servers to enhance the framework
ing to Eq. (6) as a function of the symbol duration (Tsym ) and the flexibility.
number of preamble symbols (npreamble ) and, then, the packet dura-
tion was obtained according to Eq. (7) as a function of the symbol
duration (Tsym ) and the number of header and payload symbols 5.3. Network latency test
(payloadSymbNb):
This test quantified stage-to-stage and end-to-end latencies of
TOA = Tpreamble + Theader+payload (5) the inventory-management functionality of the system for a given
configuration, since the local functionality (LoRaWAN-based) was
Tpreamble = Tsym × (4.25 + npreamble ) (6) tested in depth in Section 5.2. The metrics used to provide the
Theader+payload = Tsym × payloadSymbNb (7) results are T1 , T2 and T3 (see Table 5), which correspond to those
represented in the sequence diagram in Fig. 7. This experiment pro-
Fig. 6a shows the required TOA for 10-to-50-byte payloads for vides an overall latency approach based on the cooperation of RFID,
different SF configurations. The results range from approximately BLE and MQTT standards, with all BLE devices being programmed
70 ms to 2.8 s. Although, the number of LoRaWAN end-devices used with advertising intervals of 100 ms.
for the experiments was not sufficiently large to involve communi- In Fig. 7, T1 refers to the time elapsed from when a new product
cation bottlenecks in the R-SCM approach proposed, the results of is identified via RFID and metadata is sent via BLE to Classifica-
this test are useful to focus on a further industry-oriented scalability tion Stage, until the acknowledgement is received. T2 refers to
of the system. the time elapsed from when a new part is noticed in a Smart
For large-scale deployments, the coexistence of several LoRa Container until the BLE advertisement packet containing the inven-
devices transmitting under the same SF configuration involves tory update reaches the Hybrid Gateway. Finally, T3 is the time
10 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127

Fig. 7. Sequence diagram of the inventory-management system (T1 , T2 and T3 metrics).

Table 7 erwise, the minimum SNR threshold supported (−7.5 dB for SF7,
Results of the network latency test (in ms).
−10 dB for SF8, etc.) is considered to calculate a margin:
Metric Average Maximum Minimum STD Stages involved

T1 185.530 214.184 166.960 23.220 (i), (ii)

T2 124.635 170.744 117.920 18.631 (iii), (iv)
T3 65.724 69.909 62.106 2.457 (iv), (vii)

elapsed between an inventory update being received via BLE and

a client subscribed via MQTT to the topic-of-interest receiving the (8)
notification.
The influence of vegetation is noticed at P ♦ .1rural and P  .1rural
The measured latency of the end-to-end system is shown in
as a reduction in the SNR margin with respect to right-hand-side
Table 7, according to metrics T1 , T2 and T3 . The experimental results
locations (P ♦ .2rural and P  .2rural ), under line-of-sight. The urban
show T1 and T2 are significantly higher than T3 , which means that
measurements show the influence of infrastructures on reliability.
no communication constraints or bottlenecks might be noticed
For instance, P ◦ .2urban was measured with no surrounding build-
when scaling MQTT connectivity with the AWS IoT platform.
Metrics T1 and T2 should be analyzed according to the commu- ings, whereas in P  .3urban and P . 2urban buildings were present.
nication standards and industrial process associated with each Our experiments show that LoRaWAN presented no communica-
one. tion problems in rural scenarios, with low SFs being sufficient for
If we consider the inspection of components between the RFID- wide areas. Conversely, the urban results call for high SF allocations,
based identification of products and the classification of parts (see since only SF12 achieved 100% of success.
Fig. 1), an average latency of approximately 185 ms, T1 , would not
limit the performance of the R-SCM system. Conversely, our results 6. Conclusions
point to the fact that T2 , the BLE link between Smart Containers and
their Hybrid Gateway, could limit the speed of the system if laten- This work presents an Industry 4.0 solution for R-SCM based
cies of all stages are compared. Nevertheless, an overall minimum on a heterogeneous IoT network following DSC objectives. BLE
delay of 124 ms would not limit the inventory-monitoring func- and RFID technologies are proposed for inventory management
tionality of the system, especially considering the requirements using Smart Containers, while a LoRaWAN context network is
associated with the RL purpose of the system [40]. responsible for environmental monitoring of industrial facilities.
The network is governed by a Hybrid Gateway, responsible for
receiving BLE information and forwarding it to two back-ends: an
5.4. Geographic spread test inventory-monitoring platform in AWS (via MQTT) and a context-
information platform in TTN (via LoRaWAN). This solution was
This test evaluates LoRaWAN coverage based on RSSI and SNR developed within a real IoT prototype and applied to an industrial
(see Table 5) using the TTN Mapper [82] tool. Fig. 8 shows a map of case study regarding WEEE recovery towards CE.
the city of Albacete (Spain), the scenario used for the experiments. Four performance tests were conducted to evaluate the pro-
The gateway is represented by a cross and end-device positions (P) posal: data ingest (assessing the influence of increasing traffic),
into equidistant circles. geographical spread (wide-area communication), data size (varying
Table 8 shows and interprets the results. Since the receiver sen- payloads) and network latency (end-to-end evaluation). Our results
sitivity and, consequently, the minimum SNR is different per SF, the achieve reasonably appropriate network latencies for inventory
SNR results in Table 8 should only be compared in columns. management, suggesting that BLE communication might become
Since RSSI and SNR should not be considered absolute values, a potential bottleneck for large-scale deployments. In such sit-
Eqs. (8)a to (8)f simplify our quality (Q) interpretation criteria. Given uations, the minimum receiving time at the Hybrid Gateway
the implementation of transceivers [83], when the signal is suffi- increases proportionally with the number of Smart Containers and,
ciently strong, it is attenuated avoiding overload. RSSI and SNR are in addition, lower advertising intervals result in faster commu-
obtained after attenuation, which is why SNR values equal or higher nications at the expense of lower effectiveness. Furthermore, a
than 5 dB have been attenuated. In such cases, only RSSI provides LoRaWAN-based coverage study was carried out in Albacete (Spain)
relevant information (the lower its absolute value, the better). Oth- concerning urban and rural scenarios. Our results show how dis-
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 11

Fig. 8. Coverage study scenario in the city of Albacete with urban and rural locations.

Table 8
Geographic spread test results showing RSSI (dBm) and SNR (dB) metrics of LoRaWAN technology in urban and rural scenarios (range in meters), with a coding rate (CR) of
4/5, bandwidth (BW) of 125 kHz.

tance as well as interposed buildings or vegetation areas impact to control information accessibility depending on established rela-
on the perceived signal quality. Finally, we experimentally show to tionships between manufacturers and R-SCM operators. As a result,
what extent higher SF allocations increase the coverage area and IoT can be conceived as a tool to facilitate information manage-
discuss its implications. ment in EoL product-recovery operations contributing to the shift
The deployment and evaluation of a network prototype serve towards the CE.
as a demonstrator of the potential of IoT adoption in R-SCM sce- As future work, we plan to assess the economic and environ-
narios, which carries important managerial insights. Stakeholders mental viability of the proposed approach for different types of
could benefit from this solution achieving better management products, considering factors such as their degree-of-deterioration,
of resources based on real-time inventory information through their added-value according to condition, or the required remanu-
a low-cost and low-power infrastructure. Additionally, seamless facturing operations. Furthermore, based on the results obtained in
monitoring of facilities is achieved through a self-monitored con- this work, the IoT communications network proposed will be used
text information network able to report shop-floor incidences. With as a testbed to develop algorithms oriented to test and improve the
data-sharing being one of the barriers towards IT adoption, the energy efficiency and reliability of communications, while improv-
proposed approach enables the use of data-encryption algorithms ing the network capacity of LoRaWAN and BLE technologies in
12 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127

real industrial scenarios. For this, we aim to develop scheduling and the Committee of the Regions on a Monitoring Framework for the Circular
algorithms able to minimize collisions in the network and, thus, Economy (accessed 19.02.19) https://eur-lex.europa.eu/legal-content/EN/TXT/
?qid=1516265440535&uri=COM:2018:29:FIN#footnote25.
improve reliability through avoiding repeated transmissions. Eurostat, 2015. Your Key European Statistics: Circular Economy (accessed 19.02.19)
https://ec.europa.eu/eurostat/web/circular-economy.
Farahani, S., Otieno, W., Barah, M., 2019. Environmentally friendly disposition deci-
Declarations of interest sions for end-of-life electrical and electronic products: the case of computer
remanufacture. J. Clean. Prod. 224, 25–39.
None. Parker, D., Riley, K., Robinson, S., Symington, H., Tewson, J., Jansson, K., Ramkumar,
S., Peck, D., 2015. Remanufacturing Market Study.
Yang, X., Moore, P., Chong, S.K., 2009. Intelligent products: from lifecycle data acqui-
Acknowledgments sition to enabling product-related services. Comput. Ind. 60, 184–194.
Garcia-Mui na, F., González-Sánchez, R., Ferrari, A., Settembre-Blundo, D., 2018. The
paradigms of Industry 4.0 and circular economy as enabling drivers for the com-
This work was partially supported by the Spanish “Ministry petitiveness of businesses and territories: the case of an Italian Ceramic Tiles
of the Economy and Competitiveness” and the European Union Manufacturing Company. Soc. Sci. 7, 255.
Asif, F.M., Roci, M., Lieder, M., Rashid, A., Štimulak, M., Halvordsson, E., de Bruijckere,
(FEDER Funds) under projects ECO2016-75781-P and RTI2018-
R., 2018. A practical ICT framework for transition to circular manufacturing sys-
098156-B-C52, and the Engineering and Physical Sciences Research tems. Proc. CIRP 72, 598–602, http://dx.doi.org/10.1016/j.procir.2018.03.311,
Council (EPSRC), UK, grant no. EP/N018524/1. 51st CIRP Conference on Manufacturing Systems.
Gu, F., Ma, B., Guo, J., Summers, P.A., Hall, P., 2017. Internet of things and Big Data as
potential solutions to the problems in Waste Electrical and Electronic Equipment
References management: an exploratory study. Waste Manage. 68, 434–448.
De Toni, A., Tonchia, S., 2001. Performance measurement systems – models, charac-
Cong, L., Zhao, F., Sutherland, J.W., 2017. Integration of dismantling operations into teristics and measures. Int. J. Oper. Prod. Manage. 21, 46–71.
a value recovery plan for circular economy. J. Clean. Prod. 149, 378–386. Gunasekaran, A., Patel, C., Tirtiroglu, E., 2001. Performance measures and metrics in
Islam, M.T., Huda, N., 2018. Reverse logistics and closed-loop supply chain of waste a Supply Chain environment. Int. J. Oper. Prod. Manage. 21, 71–87.
electrical and electronic equipment (WEEE)/E-waste: a comprehensive litera- Slack, N., Chambers, S., Johnston, R., 1995. Operations Management. Pearson Educa-
ture review. Resour. Conserv. Recycl. 137, 48–75. tion.
Baldé, C.P., Forti, V., Gray, V., Kuehr, R., Stegmann, P., 2017. The Global E-waste Rajeev, A., Pati, R.K., Padhi, S.S., Govindan, K., 2017. Evolution of sustainability in
Monitor 2017 – Quantities, Flows, and Resources. United Nations University, supply chain management: a literature review. J. Clean. Prod. 162, 299–314.
International Telecommunication Union, and International Solid Waste Associ- Gunasekaran, A., Ngai, E.W., 2004. Information systems in supply chain integration
ation. and management. Eur. J. Oper. Res. 159, 269–295.
Govindan, K., Soleimani, H., Kannan, D., 2015. Reverse logistics and closed-loop sup- Srinivasan, R., Swink, M., 2018. An investigation of visibility and flexibility as com-
ply chain: a comprehensive review to explore the future. Eur. J. Oper. Res. 240, plements to Supply Chain analytics: an organizational information processing
603–626. theory perspective. Prod. Oper. Manage. 27, 1849–1867.
Dekker, R., Fleischmann, M., Inderfurth, K., van Wassenhove, L.N., 2013. Reverse Büyüközkan, G., Göçer, F., 2018. Digital Supply Chain: literature review and a pro-
Logistics: Quantitative Models for Closed-Loop Supply Chains. Springer Science posed framework for future research. Comput. Ind. 97, 157–177.
& Business Media. Govindan, K., Soleimani, H., 2017. A review of reverse logistics and closed-loop
Jerbia, R., Boujelben, M.K., Sehli, M.A., Jemai, Z., 2018. A stochastic Closed-Loop Sup- supply chains: a Journal of Cleaner Production focus. J. Clean. Prod. 142, 371–384.
ply Chain network design problem with multiple recovery options. Comput. Ind. Lembke, R., Rogers, D., 2002. Differences between forward and reverse logistics in a
Eng. 118, 23–32. retail environment. Supply Chain Manage.: Int. J. 7, 271–282.
Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M., 2013. Internet of Things (IoT): a Ashton, K., 2009. That ‘Internet of Things’ thing. RFID J. 22, 97–114.
vision, architectural elements, and future directions. Fut. Gen. Comput. Syst. 29, Da Xu, L., He, W., Li, S., 2014. Internet of Things in industries: A survey. IEEE Trans.
1645–1660. Ind. Inform. 10, 2233–2243.
Hermann, M., Pentek, T., Otto, B., 2016. Design principles for Industrie 4.0 scenarios. Boyes, H., Hallaq, B., Cunningham, J., Watson, T., 2018. The Industrial Internet of
In: 2016 49th Hawaii International Conference on System Sciences (HICSS), IEEE, Things (IIoT): an analysis framework. Comput. Ind. 101, 1–12.
pp. 3928–3937. Lu, Y., 2017. Industry 4.0: a survey on technologies, applications and open research
Zhong, R.Y., Xu, X., Klotz, E., Newman, S.T., 2017. Intelligent manufacturing in the issues. J. Ind. Inform. Integr. 6, 1–10.
context of Industry 4.0: a review. Engineering 3, 616–630. Lee, E.A., 2008. Cyber physical systems: design challenges. In: Object Oriented Real-
Helo, P., Suorsa, M., Hao, Y., Anussornnitisarn, P., 2014. Toward a cloud-based man- Time Distributed Computing (ISORC), 2008 11th IEEE International Symposium,
ufacturing execution system for distributed manufacturing. Comput. Ind. 65, IEEE, pp. 363–369.
646–656. Lee, I., Lee, K., 2015. The Internet of Things (IoT): applications, investments, and
Bittencourt, L.F., Diaz-Montes, J., Buyya, R., Rana, O.F., Parashar, M., 2017. Mobility- challenges for enterprises. Business Horizons 58, 431–440.
aware application scheduling in fog computing. IEEE Cloud Comput. 1, 26–35. Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H., Zhao, W., 2017. A survey on Inter-
Jeschke, S., Brecher, C., Meisen, T., Özdemir, D., Eschert, T., 2017. Industrial Internet net of Things: architecture, enabling technologies, security and privacy, and
of Things and Cyber Manufacturing Systems. In: Industrial Internet of Things, applications. IEEE Internet Things J. 4, 1125–1142.
Springer, pp. 3–19. Siekkinen, M., Hiienkari, M., Nurminen, J.K., Nieminen, J., 2012. How low energy is
Roberts, C.M., 2006. Radio frequency identification (RFID). Comput. Secur. 25, 18–26. bluetooth low energy? Comparative measurements with Zigbee/802.15.4. In:
Bluetooth SIG, 2014. Bluetooth 4.0, 4.1 and 4.2 Specifications (accessed 28.10.18) Wireless Communications and Networking Conference Workshops (WCNCW),
https://www.bluetooth.com/specifications/bluetooth-core-specification. 2012 IEEE, IEEE, pp. 232–237.
LoRa Alliance, 2017. LoRaWAN v1.1 Specification (accessed 28.09.18) https://lora- Dementyev, A., Hodges, S., Taylor, S., Smith, J., 2013. Power consumption analysis of
alliance.org/resource-hub/lorawantm-specification-v11. Bluetooth Low Energy, ZigBee and ANT sensor nodes in a cyclic sleep scenario.
Kneese, A.V., Ayres, R.U., d’Arge, R.C., 1971. Economics and the Environment: A In: Wireless Symposium (IWS), 2013 IEEE International, IEEE, pp. 1–4.
Materials Balance Approach. Routledge. Cho, K., Park, W., Hong, M., Park, G., Cho, W., Seo, J., Han, K., 2014. Analysis of latency
Genovese, A., Acquaye, A.A., Figueroa, A., Koh, S.L., 2017. Sustainable supply chain performance of bluetooth low energy (BLE) networks. Sensors 15, 59–78.
management and the transition towards a circular economy: evidence and some Rondón, R., Gidlund, M., Landernäs, K., 2017. Evaluating bluetooth low energy suit-
applications. Omega 66, 344–357. ability for time-critical industrial IoT applications. Int. J. Wirel. Inform. Netw. 24,
Preston, F., 2012. A Global Redesign? Shaping the Circular Economy. Chatham House 278–290.
London. Sanchez-Iborra, R., Sanchez-Gomez, J., Ballesta-Vi nas, J., Cano, M.D., Skarmeta, A.F.,
The Ellen MacArthur Foundation, 2013. Towards the circular economy. J. Ind. Ecol. 2018. Performance evaluation of LoRa considering scenario conditions. Sensors
2, 23–44. 18, 772.
Vella, K., 2016. The Circular Economy: An Investment with A Triple Win. https:// Zorbas, D., Papadopoulos, G.Z., Maille, P., Montavont, N., Douligeris, C., 2018. Improv-
ec.europa.eu/commission/commissioners/2014-2019/vella/blog/circular- ing LoRa network capacity using multiple spreading factor configurations. In:
economy-investment-triple-win en. 2018 25th International Conference on Telecommunications (ICT), IEEE, pp.
Sabbaghi, M., Esmaeilian, B., Mashhadi, A.R., Behdad, S., Cade, W., 2015. An inves- 516–520.
tigation of used electronics return flows: a data-driven approach to capture Cattani, M., Boano, C.A., Römer, K., 2017. An experimental evaluation of the reliability
and predict consumers storage and utilization behavior. Waste Manage. 36, of LoRa long-range low-power wireless communication. J. Sens. Actuator Netw.
305–315. 6, 7.
Pei, Z., Paswan, A., 2018. Consumers’ legitimate and opportunistic product return Cuomo, F., Campo, M., Caponi, A., Bianchi, G., Rossini, G., Pisani, P., 2017. EXPLoRa:
behaviors in online shopping. J. Electron. Commerce Res. 19, 301–319. extending the performance of LoRa by suitable spreading factor allocations. In:
European Parliament and Council, 2012. Directive 2012/19/EU of the European Par- Wireless and Mobile Computing, Networking and Communications (WiMob),
liament and of the Council of 4 July 2012 on Waste Electrical and Electronic IEEE, pp. 1–8.
Equipment, WEEE. Off. J. Eur. Union L 197, 38–71. Hofmann, E., Rüsch, M., 2017. Industry 4.0 and the current status as well as future
European Commission, 2018. Communication from the Commission to the Euro- prospects on Logistics. Comput. Ind. 89, 23–34.
pean Parliament, the Council, the European Economic and Social Committee
 C. Garrido-Hidalgo, T. Olivares, F.J. Ramirez et al. / Computers in Industry 112 (2019) 103127 13

Castelo-Branco, I., Cruz-Jesus, F., Oliveira, T., 2019. Assessing Industry 4.0 readiness NB-IoT, 2016. NB-IoT Standardized (accessed 17.03.18) http://www.3gpp.org/news-
in manufacturing: evidence for the European Union. Comput. Ind. 107, 22–32. events/3gpp-news/1785-nb iot complete.
Gioia, E., Passaro, P., Petracca, M., 2016. Amber: an advanced gateway solution to Sigfox, 2019. Sigfox, The World’s Leading IoT Connectivity Service (accessed
support heterogeneous IoT technologies. In: 2016 24th International Conference 27.01.19) https://www.sigfox.com/.
on Software, Telecommunications and Computer Networks (SoftCOM), IEEE, pp. The Things Network, 2019. Building a Global Internet of Things Network Together
1–5. (accessed 26.01.19) https://www.thethingsnetwork.org/.
Garrido-Hidalgo, C., Hortelano, D., Roda-Sanchez, L., Olivares, T., Ruiz, M.C., Lopez, Amazon Web Services, 2019. Amazon Web Services (accessed 26.01.19) https://aws.
V., 2018. IoT heterogeneous mesh network deployment for human-in-the- amazon.com/.
loop challenges towards a social and sustainable Industry 4.0. IEEE Access 6, ElMahgoub, K., 2014. An Encryption Algorithm for Secured Communication with
28417–28437. Passive UHF RFID Systems.
Roda-Sanchez, L., Garrido-Hidalgo, C., Hortelano, D., Olivares, T., Ruiz, M.C., 2018. Gurijala, Y.R., 2017. Performance Testing Internet of Things (accessed 26.09.18)
OperaBLE: an IoT-based wearable to improve efficiency and smart worker care https://www.infosys.com/IT-services/validation-solutions/features-opinions/
services in Industry 4.0. J. Sensors, 2018. Documents/performance-testing-iot.pdf.
Tao, F., Qi, Q., 2017. New IT driven service-oriented Smart Manufacturing: frame- ETSI, 2013. European Telecommunication Standards Institute: Compliance of
work and characteristics. IEEE Trans. Syst. Man Cybern. Syst. the SX1272/3 LoRa Modem (accessed 30.07.18) https://www.semtech.com/
Abdel-Basset, M., Manogaran, G., Mohamed, M., 2018. Internet of Things (IoT) and uploads/documents/etsi-compliance-sx1272-lora-modem.pdf.
its impact on supply chain: a framework for building smart, secure and efficient Karasiewicz, C., 2013. Why HTTP is Not Enough for the Internet of Things
systems. Fut. Gen. Comput. Syst. (accessed 01.12.18) https://www.ibm.com/developerworks/community/blogs/
Papert, M., Pflaum, A., 2017. Development of an ecosystem model for the realiza- mobileblog/entry/why http is not enough for the internet of things?lang=en.
tion of Internet of Things (IoT) services in supply chain management. Electron. Nordic Semiconductors, 2018. Cross Platform Bluetooth Low Energy Development
Markets 27, 175–189. Software (accessed 11.09.18) https://www.nordicsemi.com/eng/Products/
Wakenshaw, S., 2017. An IoT-enabled Supply Chain Integration Framework: Empir- Bluetooth-low-energy/nRF-Connect-for-Desktop.
ical Case Studies. Semtech, 2013. SX1272/3/6/7/8: LoRa Modem. Semtech (accessed 19.09.18).
Tao, M., Zuo, J., Liu, Z., Castiglione, A., Palmieri, F., 2018. Multi-layer cloud architec- The Things Network, 2018. TTN Fair Access: Limitations on Data Rate, Packet
tural model and ontology-based security service framework for IoT-based smart Size and Time on Air (accessed 27.01.19) https://www.thethingsnetwork.
homes. Fut. Gen. Comput. Syst. 78, 1040–1051. org/forum/t/limitations-data-rate-packet-size-30-seconds-uplink-and-10-
ISO, 2016. ISO/IEC 20922:2016-Message Queuing Telemetry Transport (MQTT) messages-downlink-per-day-fair-access-policy/.
v3.1.1 (accessed 01.05.18) https://www.iso.org/standard/69466.html. The Things Network, 2018. TTN Mapper (accessed 12.02.18) https://ttnmapper.org/
NFC, 2017. Near Field Communication Technology Standards (accessed 29.01.19) .
http://nearfieldcommunication.org. Semtech, 2017. LoRa Module Implementation by Semtech
Wi-Fi Alliance, 2019. Wi-Fi Certified (accessed 29.01.19) https://www.wi-fi.org/. (accessed 12.02.18) http://semtech.force.com/lora/LC Answers
ZigBee Alliance, 2019. ZigBee Alliance (accessed 29.01.19) https://www.zigbee.org. Questions?id=90644000000PmGvAAK.
LoRa Alliance, 2018. https://lora-alliance.org/ (accessed 30.04.18).