Multi-layer Fog Computing Framework for Constrained LoRa

Networks Intended for Water Quality Monitoring and Precision

Agriculture Systems

Laura García

1,2 a

, Jose M. Jimenez

1,2 b

, Sandra Sendra

, Jaime Lloret

and Pascal Lorenz

Instituto de Investigación para la Gestión Integrada de Zonas Costeras, Universitat Politècnica de València,

C/ Paranimf nº 1, Grao de Gandía – Gandia, Valencia, Spain

Network and Telecommunication Research Group, University of Haute Alsace, 34 rue du Grillenbreit, 68008,

Colmar, France

Keywords: Fog Computing, Multi-layer, Energy-saving.

Abstract: As the population of the world keeps increasing, it is necessary for the agriculture to adopt technologies that

improve the production and optimize re-sources such as water. This has been done by introducing IoT devices,

which has led to smart agriculture or precision agriculture. However, due to the remoteness of the fields, the

communication of these devices needs to be per-formed with technologies such as LoRa that has limitations

on the amount of data and the number of messages that can be forwarded. Furthermore, as there is no

connection to the electric grid, optimizing the energy consumption is a necessity. In this paper, we present a

multi-layer fog computing framework for a water quality monitoring and precision agriculture system. Data

aggregation techniques are applied at the algorithms provided for the different layers to reduce the amount of

data and the number of messages forwarded to the data center so as to improve the performance of the

constrained LoRa network and reduce the energy consumption. Furthermore, the added decision-making

provides fault-tolerance to the system if the connection to the Data Center is not available. Simulations were

performed for different functioning modes. Results show a reduction of the 80% in the amount of transmitted

data and a reduction of 85.33% in the number of for-warded messages for the most restrictive functioning

mode.

1 INTRODUCTION

The adoption of technologies that optimize

agricultural production is necessary to be able to feed

the world population, which is constantly increasing.

Furthermore, the climatological and environmental

problems threaten the future of the agricultural

production as well. The fourth industrial revolution,

Industry 4.0, has also reached agriculture resulting in

the term Agriculture 4.0 (Industry 4.0 in Agriculture:

Focus on IoT aspects., 2020). Fundamentally, it is

about applying ICT (Information and

Communication Technology) techniques to

agriculture. According to De Clercq et al. (2018), the

https://orcid.org/ 0000-0003-2902-5757

https://orcid.org/ 0000-0002-3688-7235

https://orcid.org/ 0000-0001-9556-9088

https://orcid.org/ 0000-0002-0862-0533

https://orcid.org/ 0000-0003-3346-7216

application of Agriculture 4.0 will no longer depend

so much on the use of water, fertilizers, and

pesticides, as they will be used in minimal quantities.

Furthermore, it will even be possible to cultivate in

arid areas and use clean and abundant resources such

as the sun or seawater.

Smart agriculture incorporates the contributions

of both capital and high technology, making it

possible to grow food accurately while being clean

and sustainable. For all the previously mentioned

reasons, the use of the Internet of things (IoT) applied

to agriculture is increasing day by day.

One of the biggest concerns, when deploying IoT

devices is their power consumption. Currently, when

García, L., Jimenez, J., Sendra, S., Lloret, J. and Lorenz, P.

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture Systems.

DOI: 10.5220/0010618300460055

In Proceedings of the 18th International Conference on Wireless Networks and Mobile Systems (WINSYS 2021), pages 46-55

ISBN: 978-989-758-529-6

applied to agriculture, most of the time they are

installed in areas where it is not possible to connect

the device to the electricity grid. For this reason,

devices and technologies with minimal energy

consumption are often used. To achieve this, in the

field of communications it is ideal to use technologies

such as Low-Power Wide-Area Network (LPWAN).

Among the LPWAN technologies, we can

highlight Long Range (LoRa) (LoRa Alliance, 2020),

due to its low cost, long range, and the use of license-

free bands. The main problem that can be found is that

the information transmitted through the air must be as

optimized as possible to reduce as much as possible

what is known as "time on air" or "airtime", that is,

the time necessary to transmit the message from the

sender node to the receiver (gateway). The more time

utilized for data transmission, the greater the

saturation of the frequency and greater energy

consumption; therefore, it is vital to keep the payload

of the transmissions as low as possible.

Greenfield defines and differentiates Fog

Computing and Edge Compu-ting. When using Fog

Computing a decentralized network structure is

employed in which resources, including data and

applications, will be found somewhere between the

data source and the Cloud (Fog Computing vs. Edge

Computing: What’s the Difference?, 2016). By using

Edge Computing, intelligence is brought into

individual hardware systems such as sensors. Using

Edge Compu-ting, the source devices are already in

charge of filtering data. Redundant data and even

false positives can be removed depending on the

architecture.

In our work we present the proposal of a multi-

layer fog computing framework for a precision

agriculture and irrigation water quality monitoring

system. Which has been designed to be implemented

in two areas. One of the control zones are the canals

that transport the irrigation water, where the salinity

or turbidity levels of the water are observed, and if

any of the two parameters exceeds a threshold, an

alarm is forwarded. The other zone is that of the crop

fields, where parameters such as soil moisture and

soil salinity will be observed. Through our proposal,

decision-making, regarding the actions to be carried

out to achieve optimal irrigation of the crops, will

occur in nodes located near the sensor nodes.

Furthermore, data aggregation is performed to adapt

to the constrained LoRa network. In this way, the

energy consumption will be reduced by filtering the

data and reducing the number of messages sent to the

data center. The proposed framework will provide the

system with fault tolerance capabilities as well,

eliminating the need of depending on continued

computations at the Data Center.

The remainder of this paper is organized as

follows. Section 2 presents the related work. The

multi-layer fog computing framework description is

explained in Section 3. The simulation results are

carried out in Section 4. Finally, Section 5 draws the

main conclusions and future works.

2 RELATED WORK

The introduction of several layers to the edge and fog

computing architecture provides many benefits

compared to the classic cloud architecture. Gia et al.

(2019) discuss these benefits applied to smart systems

in remote areas. At these areas, the chosen

communication technology, such as LoRa, has a low

bandwidth and performing the analysis and decision-

making activities at the edge allows providing more

functionalities. The presented architecture has an

edge, a fog and a cloud layer and performs image

compression based on CNNS. Results show 67% of

data size reduction with less than 5% of

decompression errors. Guardo et al. (2018) also

presents a fog computing framework with two tiers

intended for precision agriculture. The first tier is

comprised of the sensing nodes and the second tier is

the gateway. Both tiers perform data filtering and

analysis to reduce the amount of data forwarded to the

Cloud. The LoRa and MQTT protocols were utilized

for communication. The authors expected a reduction

in cost, a waiting times and load balancing as a result

of implementing the proposed architecture.

On the other hand, conventional fog computing

networks have some challenges as well. Chang et al.

(2017) present a fog computing infrastructure called

In-die Fog in order to solve some of these challenges.

With Indie Fog, the authors aim to provide a solution

that can be implemented with consumer devices

eliminating the need and the restrictions of the

devices owned by the service provider. Furthermore,

fog computing reduces latency and provides

communication and computation efficiency. Indie fog

can be implemented in an integrated manner, where

the router incorporates a virtual machine to perform

computations, or in a collaborative manner, where a

computer is connected to the router. However, other

devices such as smartphones or vehicles can also be

used as fog devices. Wang et al. (2019) designed a

multilayer system for edge computing. The

architecture is comprised of three layers being the

edge device, the Access Point and the Cloud Center.

The authors divide the system into a blocking and a

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture

Systems

nonblocking state. Results showed a minimization of

recovery time for the block-ing state and reduced

latency for the nonblocking state. Another multi-tier

architecture was presented by Chekired and Khoukhi

(2018). The authors also present the Simulated

Annealing Algorithm to determine the best allocation.

A probabilistic model is utilized to determine and

improve the efficiency of the presented architecture.

The simulation results show that a reduction of 35%

in response time can be obtained utilizing the 3-tier

architecture instead of the flat architecture. For the

case of the 2-tier topology, a 20% reduction was

obtained. Regarding offloading, a reduction of 30%

was obtained for the 3-tier topology, outperforming

other designs. Lastly, an improvement in

performance was obtained as well utilizing multi-tier

architectures.

Although some work has been done in (Guardo et

al., 2018) regarding fog and edge computing in

agriculture, it only considers the fields and a small

number of devices. In this paper, the concept is

extended and applied to both the fields and the canals

that provide water to the fields so as to determine the

quality of the water. Further-more, data aggregation

algorithms are provided for the layers of the

framework that perform fog computing.

3 MULTI-LAYER FOG

COMPUTING FRAMEWORK

DESCRIPTION

In this section, our multi-layered fog computing

proposal for a precision agriculture and irrigation

water quality monitoring system is presented.

The deployment of the water quality monitoring

system for irrigation purposes and the precision

agriculture system is presented in Figure 1. The Canal

Area is comprised of a series of subcanals in a comb

shape where the biosorption pro-cess for water

purification is performed. The Field Area is where the

fields are situated. These fields can be further divided

into different zones to apply different processing to

each of the areas and have a more detailed overview

of the state of the plants. The Urban Area is the zone

where the Gateway is located and connected to the

Data Center. There are clusters of sensing nodes at

each subcanal and deployed on each zone of the

fields, with a cluster head for each cluster. Actuator

nodes are deployed as well at each subcanal to

manage the gates to the biosorption process and to

control the amount of water for the irrigation of the

fields. Furthermore, the cluster head forwards the data

to the aggregator of their Area. Then, the aggregator

forwards the data to the Gateway destined to the Data

Center. The CH nodes and Actuators at each Zone are

detailed at Table 1. The system is scalable, and more

zones can be added when necessary. If monitoring

more fields is necessary, more gateways could be

added so that the data transmission is divided into the

deployed gateways and a bottleneck scenario is

avoided.

The architecture of the system is presented in

Figure 2. The architecture is com-prised of the

following layers:

• Layer 1: The first layer is the layer comprised

of the sensor and actuator nodes. These devices

forward all data to the next layer and do not

perform any computations. The data

acquisition is performed at different

frequencies depending on the selected settings.

These settings are the Research Mode, the

Advanced Farmer Mode and the Regular

Farmer Mode. Table 2 presents the

characteristics of each mode. For the case of the

actuators, the actions are performed when the

message with the new action is received. Then,

a message with the new state of the actuator is

forwarded to the data center.

• Layer 2: This layer is comprised of the Cluster

Head nodes. These nodes receive the data from

the sensing nodes and perform the first data

aggregation process detailed by the Algorithm

1. For the Canal Area, these nodes evaluate the

quality of the water by considering the values

obtained by all the nodes at the cluster. The CH

node receives the turbidity or salinity levels

from the sensor nodes of the cluster. The outlier

values are discarded. This is performed by

calculating the variance 𝜎





and determining if

it surpasses the 𝑇ℎ



threshold. Then, the

CH compares all the received values using the

𝐶



variable to obtain a final result of

the salinity or turbidity levels of the water. If

the turbidity or salinity surpasses the

𝑇ℎ



threshold, an alarm is forwarded to

the next layer. For the case ofthe CH nodes of

the field area, the same process is performed

with soil moisture and soil salinity. That way,

the Aggregator nodes at layer 3 and the Data

center are able to determine if the variations in

the calculations due to water stress or salinity

should be applied when determining the

necessary amount of water for irrigation.

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

• Layer 3: This layer is comprised of the

Aggregator nodes. These nodes receive the data

from each of their areas, being the canal area or

the field area. This node performs the decision-

making process that determines the actions of

the actuator of their areas. At the canal area, the

Aggregator node sends a message to the

actuator node of the specific canal to open or

close the gates. At the field Area, the

Aggregator node calculates the water

requirements for a time frame of one month and

forward the actions to the Actuator when

irrigation is required. The data exceeding the

time frame of one month is deleted from the

storage system of the Aggregator node. The

data is aggregated as detailed in Algorithm 2

and forwarded to the data center when the data

forwarding timer is reached so the user can

access the history of all the variables.

According to the state of the actuators in the

Canal Area and the selected settings, the

amount of data forwarded to the data center

varies.

• Layer 4: This layer is comprised of the

gateway. This node will store all the data in the

rare case the Data Center or the connection to

the data center is down. This layer does not

perform data aggregation nor performs

computing as all the necessary data aggregation

and computation has been performed on the

lower layers.

• Layer 5: This layer is comprised of the Data

Center. The data center stores all the data and

processes the information to perform analysis

and predictions on water quality, water

requirement and quality of the soil, among

others.

By providing a multi-layered fog computing

functionality to the topology of the water quality

monitoring and precision agriculture system, the

obtained benefits are twofold. On the one hand, it

provides the system with fault-tolerance capabilities

by providing autonomy to each of the layers of the

architecture in case any of the elements of the

network gets damaged or stops functioning and the

decision made at the Data Center cannot be

forwarded. This is a key aspect considering the tree

topology of the system. On the other hand, it reduces

the energy consumption by filtering the data and

limiting the number of messages that are forwarded

to the data center. This reduction in data and

messages also helps to reduce the collisions that may

be caused when different LoRa nodes transmit at the

same time. It also allows the system to meet the duty

cycle requirements of LoRa. However, it is important

to consider that in our proposal, a multi-hop LoRa

network is considered thus protocols such as the ones

in (Liao et al., 2017) are utilized instead of

LoRaWAN. While the use of WiFi results in more

energy consumption than using communication

technologies with similar coverage such as ZigBee,

WiFi is often used due to its convenience,

accessibility and low price of the devices.

Nonetheless, the presented proposal would lead to a

reduction in energy consumption if ZigBee was

utilized as in (Truong et al., 2021).

Figure 1: Water quality monitoring and precision agriculture system.

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture

Systems

Table 1: CH nodes and Actuator nodes at each Zone.

Zone CH Actuator

Main canal CH 1 Canal Area Actuator Node 1 &2

Secondary Canal 1 CH 2 Canal Area Actuator Node 3 &4

Secondary Canal 2 CH 3 Canal Area Actuator Node 5 & 6

Secondary Canal n CH n+1 Canal Area Actuator Node 2n+1

Biosorption Output CH n+2 Canal Area Actuator Node 2n+2

Field Area Zone 1 CH 1 Field Area Actuator Node Field Area 1

Field Area Zone 2 CH 2 Field Area Actuator Node Field Area 2

Field Area Zone m CH m Field Area Actuator Node Field Area m

Figure 2: Multi-layered fog architecture of the precision agriculture and irrigation quality monitoring system.

Table 2: Data forwarding and aggregation settings.

Settings

Data Acquisition

Frequency

Data Aggregation

Data Forwarding

Frequency

Research Mode 10 minutes Data driven aggregation at layer 2 4 times a day + Alerts

Advanced Farmer Mode 30 minutes

Data driven aggregation at layer 2

and layer 3

2 times a day + Alerts

Regular Farmer Mode 1 hour

Data driven aggregation at layer 2

and layer 3

Once a day + Alerts

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

lgorithm 1: Data Aggregation at layer 2.

)

Gather data from the sensors of the CH node

2) Receive the data from the n sensors of the cluste

)

for each variable var do

4) for each node i in the cluster do

𝜎





∑

(







)









if 𝜎





>𝑇ℎ







then

)

Discard data from sensor i

8) end if

If 𝑉



> 𝑇ℎ



then

10)

𝐶



= 𝐶



11) end if

)

end for

13)

if 𝐶



≥

3𝑛

 then

14)

Add the average of the value for the variable 𝑉





to Alert-Payload string

15)

else if 𝑉





>𝑇ℎ



then

16)

Add the average of the value for the variable 𝑉



to the Alert-Payload string

)

else

18)

Store aggregated data 𝑉



)

end if

20) end for

)

if there is data on the

lert-Pa

load strin

then

22) Send Alert message with the content of

lert-Payload

)

end if

24) if timer for data forwarding has been reache

then

)

Forward stored a

ated data

26) end if

)

End

4 SIMULATION RESULTS

In this section, the results of the simulations for

amount of forwarded data and number of forwarded

packets are presented. As the collision management

is not part of the scope of this paper, it is assumed that

there are no collisions.

The Canal area is comprised of four clusters that

forward the data to one aggregator node. There are

eight actuator nodes in this area. For the Field area,

three zones were considered with three clusters per

zone, one aggregator node which is an

agrometeorological station as well, and three actuator

nodes. The sensing nodes and the CH nodes

communicate through WiFi. The CH nodes and the

Aggregator nodes communicate through LoRa at the

EU 863-870 frequency band, with a bandwidth of 125

kHz and a spreading factor of SF8. This LoRa settings

allow a maximum payload of 222 Bytes including the

LoRa header.

The data forwarded by each layer to the next layer

of the hierarchy on the Regular Farmer Mode is

presented in Figure 3 a) and b). The algorithms allow

reducing substantially the amount of data forwarded

by the higher Layers to the Data Center with an 83%

for layers 1 and 2 and an 80% when the data

forwarding timer is reached. Furthermore, a reduction

of 69% in the transmitted data at the forwarding time

is obtained compared to not performing data

aggregation. This reduction in the forwarded data

leads to a reduction in the energy consumption of the

devices as the higher energy consumption is produced

when data is transmitted.

At the Advanced Farmer Mode, as it can be seen

in Figure 3 c) and d), the amount of forwarded data is

decreased due to the data acquisition frequency of 30

minutes. The state of the Actuator nodes keeps being

forwarded each hour. Thus, there is a fluctuation in

the amount of data forwarded each time the timer is

reached. Furthermore, the data is aggregated at Layer

3 as well and the data is forwarded to the data center

twice a day. With this mode, compared to not per-

forming data aggregation, a reduction of 60% was

achieved when the forwarding time is achieved. With

data aggregation, a reduction of 66% of the forwarded

data was obtained compared to the Research mode.

The reduction reached a 76% at the times where the

states of the actuator are forwarded.

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture

Systems

lgorithm 2: Data Aggregation and decision-making at layer 3.

)

date Actuator decision makin

rules from the Data Cente

2) Receive data from the devices in layer 2.

)

Receive Actuator State

4) if Canal Area Alert received then

5) Forward Action message to the required actuator so as to open the gates of the biosorption canal closest to the

CH node that activated the alar

)

Forward Alert messa

e destined to the Data Cente

7) end if

)

if Field Area Alert receive

then

9) Store Alert for further processing

)

Forward Alert Messa

e destined to the Data Cente

11) end if

)

if data

orwardin

time

is reached then

13) if Research Mode then

)

if all

ates are closed then

15) Add the data from the Main Canal to the Payload string

)

else

17) Add data from all CH nodes at canal area to the Payload string

)

end if

19) Add data of the CH nodes at the field area to the Payload string

)

else if Farmer Mode then

21) Add data of the Main Canal and the Biosorption Output to the Payload

)

for each field area do

23) for each Zone at Field area do

)

Calculate avera

e of the variables measured b

all CH nodes at the same zone

25) end for

)

end for

27) Add average data of the field zones to the Payload

)

end if

29) Forward message with the data stored at the Payload string

)

end if

31) If water_requirement_calculation_time

is reache

then

)

For each zone in field area do

)

if water

stress alert && Salinit

alert receive

then

34) Calculate water requirements with wate

-stress and salinity modifications

)

else if water

stress alert receive

then

36) Calculate water requirements with wate

-stress modifications

)

else if Salinit

alert receive

then

38) Calculate water requirements with salinity modifications

)

else

40) Calculate water requirements

)

end if

42) if irrigation_day is true then

43) Forward Action message to the actuator nodes of each area with the amount of water needed for the next

irrigation

)

end if

45) end for

)

end if

47) End

The Researcher mode performs the data acquisition

process each 10 minutes and thus, the high amount of

forwarded data (See Figure 3 d) and f)). As it can be

seen, the nodes in Layer 1 forward all the data to the

CH nodes in Layer 2. The state of the Actuator nodes

is forwarded each hour. The Researcher Mode obtains

a reduction of 35% at the forwarding times compared

to not performing any data acquisition. In this case,

no data aggregation was performed at Layer 3.

Furthermore, there is a water salinity alarm on the

first day. At the times the data forwarding timer is

reached, the data forwarded nearly reaches 20000

Bytes for an hour.

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

Regarding the number of forwarded messages, for

the Regular Farmer mode, the number of forwarded

messages remains between 39 and 55 messages per

hour (See Figure 4 a) and b). With an 85.33% of

reduction at peaks compared to the Researcher Mode

and 29% compared to not performing data

aggregation. This mode is optimal for remote areas as

it would allow other LoRa settings with more re-

strictions regarding the number of messages that can

be forwarded by each de-vice. Furthermore, it would

not be detrimental to the farmer as the calculations of

the irrigation requirements only need one measure of

each variable per day except for the meteorology data

were the maximum and minimum values of

temperature and relative humidity are necessary.

At the Advanced Farmer Mode, the number of

messages oscillates between 78 and 83 messages per

hour except for the first hour with 53 messages (See

Figure 4 c) and d)). The reduction of messages at the

peaks is 68,67% compared to the Researcher Mode

and 21% compared to not performing any data

aggregation.

The results of the Researcher Mode are presented

in Figure 4 e) and f)). Between 234 and 249 messages

are forwarded per hour except when the system is

firstly activated where 53 messages are generated. As

it can be seen, the hierarchical structure of the

framework allows reducing the number of forwarded

messages at each layer.

Figure 3: Data forwarded by each layer at the a) Regular Farmer Mode without data aggregation, b) Regular Farmer Mode

with data aggregation, c) Advanced Farmer Mode without data aggregation, d) Advanced Farmer Mode with data aggregation,

e) Researcher Mode without data aggregation and Researcher Mode with data aggregation.

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture

Systems

Figure 4: Number of messages forwarded by each layer at the a) Regular Farmer Mode without data aggregation, b) Regular

Farmer Mode with data aggregation, c) Advanced Farmer Mode without data aggregation, d) Advanced Farmer Mode with

data aggregation, e) Researcher Mode without data aggregation and Researcher Mode with data aggregation.

Furthermore, a reduction of 9% of the messages at

the peaks compared to not performing data

aggregation was achieved. This is important

regarding LoRa as there is a limitation in the number

of messages that can be forwarded due to the duty

cycle. Other LoRa settings would not support the

Researched Mode. This is a key aspect if longer

distances need to be reached as higher spreading

factor values would be necessary and thus, the

number of messages that could be forwarded would

decrease.

5 CONCLUSION AND FUTURE

WORK

The introduction of IoT technologies in agriculture

has led to the optimization of the food production and

resources such as water. However, as precision

agriculture systems often need to be deployed on

remote areas, technologies such as LoRa must be

employed. But these technologies introduce

restrictions on the amount of data that can be

forwarded and the number of messages that can be

transmitter per device due to the duty cycle.

Furthermore, the impossibility of connecting the

devices to the grid also introduces energy

consumption constrictions, which is tied to the

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

performed data transmissions. In this paper, a multi-

layer fog computing framework for water quality

monitoring and precision agriculture has been

presented.

Two algorithms for data aggregation and

decision-making have been provided as well. These

algorithms also provide fault-tolerance to the network

providing certain levels of autonomy in case the

connection with the Data Center is severed.

The simulation results show a reduction in the

amount of forwarded data for the Advanced Farmer

Mode and the Regular Farmer Mode of 66% and 83%

respectively. Furthermore, a reduction of 68.67% in

the number of messages was obtained for the

Advanced Farmer Mode and of 85.33% for the

Regular Farmer Mode. These results also lead to a

reduction in energy-consumption as the most energy

is consumed when transmitting data.

As future work, other forms of gateways so as to

add another layer of fog computing in networks that

present other types of constrictions will be considered

such as the use of drones for precision agriculture as

a mobile gateway like in (García et al., 2020).

ACKNOWLEDGEMENTS

This work has been supported by European Union

through the ERANETMED (Euromediterranean

Cooperation through ERANET joint activities and

beyond) project ERANETMED3-227

SMARTWATIR, and by the Universitat Politècnica

de València through the Program "Convocatoria 2020

de contratación de Doctores para el sistema español

de Ciencia, Tecnología e Innovación, en Estructuras

de Investigación de la Universitat Politècnica de

València (PAID-10-20)".

REFERENCES

Chang, C., Narayana Srirama, S., & Buyya, R. (2017). Indie

Fog: An Efficient Fog-Computing Infrastructure for the

Internet of Things. Computer, 50(9), 92–98.

https://doi.org/10.1109/mc.2017.3571049

Chekired, D. A., & Khoukhi, L. (2018, May). Multi-Tier

Fog Architecture: A New Delay-Tolerant Network for

IoT Data Processing. 1–6.

De Clercq, M., Vats, A., & Biel, A. (2018, February).

Agriculture 4.0: The Future 0f Farming Technology.

World government summit. https://www.world

governmentsummit.org/api/publications/document?id=

95df8ac4-e97c-6578-b2f8-ff0000a7ddb6

Fog Computing vs. Edge Computing: What’s the

Difference? (2016, September 22). Automation World.

https://www.automationworld.com/products/data/blog/

13315784/fog-computing-vs-edge-computing-whats-

the-difference

García, L., Parra, L., Jimenez, J. M., Lloret, J., Mauri, P. V.,

& Lorenz, P. (2020). DronAway: A Proposal on the Use

of Remote Sensing Drones as Mobile Gateway for

WSN in Precision Agriculture. Applied Sciences,

10(19), 6668. https://doi.org/10.3390/app10196668

Gia, T. N., Qingqing, L., Queralta, J. P., Zou, Z., &

Westerlund, T. (2019, September). Edge AI in Smart

Farming IoT: CNNs at the Edge and Fog Computing

with LoRa. IEEE AFRICON, Accra, Ghana.

Guardo, E., Di Stefano, A., La Corte, A., Sapienza, M., &

Scatà, M. (2018). A Fog Computing-Based IoT

Framework for Precision Agriculture. Journal of

Internet Technology, 19(5), 1401–1411.

Industry 4.0 in Agriculture: Focus on IoT aspects. (2020,

July). https://ec.europa.eu/growth/tools-databases/dem/

monitor/sites/default/files/DTM_Agriculture%204.0%

20IoT%20v1.pdf

Liao, C. H., Zhu, G., Kuwabara, D., Suzuki, M., &

Morikawa, H. (2017). Multi-Hop LoRa Networks

Enabled by Concurrent Transmission. IEEE Access, 5,

21430–21446.

https://doi.org/10.1109/access.2017.2755858

LoRa Alliance. (2020). What is LoRaWAN® Specification.

https://lora-alliance.org/about-lorawan

Truong, V. T., Nayyar, A., & Ahmad Lone, S. (2021).

System Performance of Wireless Sensor Network

Using LoRa–Zigbee Hybrid Communication.

Computers, Materials & Continua, 68(2), 1615–1635.

https://doi.org/10.32604/cmc.2021.016922

Wang, P., Yao, C., Zheng, Z., Sun, G., & Song, L. (2019).

Joint Task Assignment, Transmission, and Computing

Resource Allocation in Multilayer Mobile Edge

Computing Systems. IEEE Internet of Things Journal,

6(2), 2872–2884. https://doi.org/10.1109/jiot.2018.28

76198

Multi-layer Fog Computing Framework for Constrained LoRa Networks Intended for Water Quality Monitoring and Precision Agriculture

Systems