a multidimensional internet of things testbed system

Research ArticleA Multidimensional Internet of Things Testbed System:Development and Evaluation

Ibrahim S. Alsukayti

Department of Computer Science, College of Computer, Qassim University, Buraydah, Saudi Arabia

Correspondence should be addressed to Ibrahim S. Alsukayti; [email protected]

Received 4 August 2020; Revised 5 September 2020; Accepted 25 September 2020; Published 14 October 2020

Academic Editor: Shuhui Yang

Copyright © 2020 Ibrahim S. Alsukayti. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The technological breakthrough of the Internet of Things (IoT) drives the emergence of a wide scope of smart IoT solutions indifferent domains. Advancing the different technological aspects of these solutions requires effective IoT implementations andexperimentations. This is widely addressed following low-cost and scalable methods such as analytical modeling and simulation.However, such methods are limited in capturing physical characteristics and network conditions in a realistic manner.Therefore, this paper presents an innovative IoT testbed system which facilitates practical experimentation of different IoTsolutions in an effective environment. The testbed design was developed towards a general-purpose multidimensional support ofdifferent IoT properties including sensing, communication, gateway, energy management, data processing, and security. Theimplementation of the testbed was realized based on integrating a set of robust hardware components and developing a numberof software modules. To illustrate its effectiveness, the testbed was utilized to experiment with energy efficiency of selected IoTcommunication technologies. This resulted in lower energy consumption using the Bluetooth Low Energy (BLE) technologycompared to the Zigbee and 6LoWPAN technologies. A further evaluation study of the system was carried out following theTechnology Acceptance Model (TAM). As the study results indicated, the system provides a simple yet efficient platform forconducting practical IoT experiments. It also had positive impact on users’ behavior and attitude toward IoT experimentation.

1. Introduction

Internet of Things (IoT) is a technological move towardseffective convergence of the physical and digital worlds. Itstarts to significantly influence our daily lives as it finds itsway into different governmental, industrial, and commercialdomains. Nowadays, there is a growing interest in IoT devel-opment considering different applications such as smarthome, e-healthcare, and intelligent transportation systems.Real deployment of IoT systems becomes widespread in dif-ferent regions of the globe. The number of IoT-enabledobjects is increasing to reach high figures in the near future.

Along with the rapid development of the IoT technology,wide range of IoT solutions has emerged. On the other hand,advancing the different technological aspects of IoT becomesa trend research direction. Different research questions havebeen raised up in a broad array of IoT applications. These arerelated to different aspects such as which and how IoT com-munication technologies should be set up, what and how IoT

data should be collected and processed, and where IoT nodesshould be positioned. However, such considerations criticallyrequire effective testing and experimentation.

There are a number of methods that can be adopted forthis process. The most common ones are analytical model-ing, simulation, and testbed experimentation. Analyticalmodeling relies on mathematical and statistical methods,whereas simulation is widely adopted as a low-cost and flex-ible experimentation approach. However, these methodshave well-known limitations as they rely on modeling physi-cal characteristics and reflecting physical phenomena usingapproximations and simplifications. This makes it complexto capture all natural hardware characteristics and realisticnetwork conditions. It is challenging for simulation softwareto efficiently well-handle imperfections in radio communica-tion, hardware interaction, sensing, and network traffic.Therefore, the need for IoT implementation and evaluationin a natural and real environment is evident. Having experi-mental IoT setups developed using physical IoT components

HindawiWireless Communications and Mobile ComputingVolume 2020, Article ID 8849433, 17 pageshttps://doi.org/10.1155/2020/8849433

https://orcid.org/0000-0002-6925-598X

https://creativecommons.org/licenses/by/4.0/

https://creativecommons.org/licenses/by/4.0/

https://doi.org/10.1155/2020/8849433

with real IoT data traffic is a more effective experimentationapproach. Although it comes with a cost in terms of moneyand time, building IoT testbed systems would allow increas-ing the realism of testing environment and improving thecredibility of the evaluation process.

On the other hand, building a motivating academic andresearch environment is important for advancing the processto a further limit. This is one of the main strategic objectivestowards which the College of Computer in Qassim Univer-sity strives. Accordingly, a number of academic resourceshave been made available and a variety of research activitieshas been supported. However, there are still no customizedfacilities available to support IoT research and education inthe college. Researchers usually have to build their ownexperimental IoT setups for carrying out certain implemen-tation and testing procedures. Educators have limitedresources to acquire hands-on experience and interactivelearning when it comes to practical IoT concepts. Given thegrowing interest in IoT technology, having a general-purpose and open-access IoT testbed would provide collabo-rative research environment and interactive learning plat-form. It can be, for example, utilized to investigate IoTcommunication technologies, study IoT network perfor-mance, test IoT security techniques, prototype IoT applica-tions, and analyze IoT data. Developing an IoT testbed isenvisaged to open the door for different opportunitiesincluding multidisciplinary cooperation. It is a cost-effectiveand time-saving approach to effectively promote IoTresearch and education.

Such a challenge is addressed in this paper which pre-sents the design and implementation of a practical testbedsystem for IoT experimentation. It represents a step forwardtowards moving IoT experimentation to a more realistic andpractical level. The main contribution of the proposedtestbed system is the novel multidimensional support provid-ing the following:

(i) An IoT experimentation facility for effective testingof different IoT aspects in a realistic and controlledenvironment. It can be practically utilized to experi-ment with IoT networking considering a number ofIoT communication protocols. It also supports IoTexperimentation targeting IoT data processing, gate-way operations, cloud integration, node deployment,and security

(ii) An educational platform for practical and interac-tive IoT teaching-learning process. It can be easilyimplemented for practical teaching of differentIoT concepts and technologies in addition toenabling interactive and practice-based IoTeducation

The testbed also features a novel combination of a num-ber of characteristics including the following:

(i) General-purpose multidimensional support of dif-ferent IoT properties (sensing, communication,gateway, energy management, data processing,and security)

(ii) Scalable, portable, and simple system architecture

(iii) Cost-effective design and implementation usingCommercial Off-The-Shelf (COTS) components

(iv) Effective support of a wide range of IoT communi-cation technologies (BLE, Zigbee, 6LoWPAN, andLoRaWAN)

(v) Easy integration of different sensor devices

(vi) Cloud integration

(vii) Effective IoT data visualization

(viii) Flexible access to the testbed using a web interfaceand API calls

(ix) Easy implementation of a wide range of IoTapplications

(x) Support of node mobility

The testbed was built based on effective design consid-ering high levels of simplicity, scalability, usability, porta-bility, and cost-effectiveness. The main focus was on therealization of a ready-to-use and easy-to-control multidi-mensional testbed for time-saving and cost-cutting IoTexperimentation. The testbed implementation was realizedbased on integrating a set of well-known hardware compo-nents and developing a number of software modules. Theefficiency of the testbed to implement and run differentIoT experiments is illustrated in this paper by a simpleIoT use case. The evaluation of the system was conductedusing the Technology Acceptance Model (TAM). As theresults indicated, the system provides simple-to-use andefficient facility for effective IoT testing. It is also evidentthat the system succeeded in promoting positive attitudestowards practical IoT experimentation and attractive facil-ity for future IoT testing.

In the following section, a summary of the related work isprovided. Section 3 presents a technical overview of the IoTtechnology is presented. In Section 4, the main requirementsof the proposed testbed system are discussed. Sections 5 and6 present the system design and implementation of the pro-posed IoT testbed, respectively. Section 7 illustrates someexample use cases of the testbed. A description of the systemevaluation and discussion of the results are presented in Sec-tion 8. Section 9 is the conclusion to this paper.

2. Related Work

The research community has realized the evolving IoT tech-nology and recognized the need for effective IoT experimen-tation. As indicated in [1], it is a common research practicefor the simulation-based IoT studies to be validated withphysical testbed experimentation. There have been a numberof research efforts made to develop and deploy differentphysical IoT experimentation testbeds. Most of these testbedswere made publically available and openly accessible withvarying facilities and functionalities. This has led to the emer-gence of new concepts in regard to IoT experimentation.

2 Wireless Communications and Mobile Computing

These include the IoT Testbed-as-a-Service (TaaS) [2] andExperiment-as-a-Service (EaaS) [3].

In [4], a small-scale testbed with cost-effective and easy-to-build design was presented. It was deployed at the buildingof Electrical Engineering Department of the University ofNorth Texas. The design of the testbed was based on a singlebase station and a set of sensor nodes running the Zigbeeprotocol and having different sensors. In addition, a numberof software components for sensor data transmission, gate-way functionality, database management, and user interfacewere implemented.

The Twonet testbed was proposed in [5] with a large-scale design. It was deployed throughout a university build-ing of four floors. The main objective of Twonet is supportingmultichannel wireless sensor networking. The design ofTwonet is based on three-tier architecture. At tier 1, a LinuxPC server exists to control testbed access and experimenta-tion through a custom web interface. It also has a set of 20Raspberry Pi nodes, interconnected using PoE switch, asproxies at tier 2 for collecting debug logs and sensor nodedata. Each proxy interconnects a set of sensor nodes (at tier3) implemented using Opal sensor devices.

Another large-scale testbed is SmartCampus [6] locatedin the Centre for Communication Systems Research (CCSR)building at the University of Surrey, UK. It was developed aspart of the European SmartSantander experimental facility[7]. It was deployed across an entire building of three floorsin a real-world environment with realistic operational condi-tions. The testbed had a user-centric design which enableseffective study of user behavior in IoT environments andthe efficiency of loT solutions. The testbed was designed fol-lowing a 3-tier model connecting a number of IoT nodes todifferent gateways to be then connected to a cloud. EachIoT node was implemented using a TelosB mote with differ-ent sensors and a IEEE 802.1 5.4 radio module. The testbedalso incorporates 30 Android Smartphone carried by theusers and connected to the gateways. The testbed relies on amanagement framework for configuring and controlling itsfunctionalities. It provides a user interface enabling resourcediscovery and reservation, topology display, experiment con-figuration and execution, and data analysis.

In [8], the testbed was built with a small-scale setup anddesigned for different IoT application domains. The mainobjective of the testbed is providing a low-cost, reliable, andcustomizable testbed. It enables low power consumption, testrepeatability, and code portability. The testbed design isbased on a number of sensor nodes wirelessly connected adedicated gateway using nRF24L01+ chipsets with low powerconsumption. The sensor nodes and gateway were imple-mented using Arduino UNO and Arduino DUE, respec-tively. The gateway is connected to a server which runsthree software components: Python adapter server, Apacheweb server, and MySQL database server.

Supersensors were presented in [9] as a campus-widetestbed for the University of Glasgow. It was based on a dis-tributed data collection system with flexible and scalablemicrocomponent design architecture. The testbed can beused for implementing different use cases for smart campusdevelopment. The testbed architecture is composed of a

number of sensor nodes (Raspberry Pi with a set of sensors)connected to a central server. The sensor nodes run differentdaemons for real-time sensor data collection and networkconnectivity establishment. At the server side, a publish/sub-scribe model with a queuing system is implemented for effec-tive real-time data processing and storage in a database. TheProtocol Buffers language (proto3) is also used for serializingand parsing data. In addition, a heartbeat mechanism isimplemented in the server to enable remote health and avail-ability monitoring as well as reprogramming and configura-tion of the sensor nodes. The server provides a userinterface allowing users to access node and sensor data usingREST APIs.

In [10], another IoT testbed deployed at Okayama Uni-versity of Science across a five-floor academic building waspresented. The testbed was constructed using a set of NotePCs and Raspberry Pi devices. A Distributed Internet TrafficGenerator software was implemented to generate differenttypes of network data traffic in the testbed. It was also usedto obtain information about different metrics in regard todata transmission. The testbed enables effective evaluationof IoT protocols and algorithms considering different realis-tic networking scenarios and parameters.

In [11], an IoT testbed was developed for dependabilitycompetition to quantitatively test low-power wireless sensingsystems in environments with high radio interference. Thetestbed was deployed using a number of TelosB devices withlow-power wireless radios in a building at Graz University ofTechnology in Austria. A Raspberry Pi running a controlsoftware is interfaced with each TelosB device using anopen-hardware interface board capable of energy consump-tion monitoring. Users can load their binary images intothe boards to run IoT experiments for certain duration. Thesystem provides different performance measures such asdelay, consumed energy, and reliability.

An IoT testbed developed following a minimalisticapproach to maintain cost-effective implementation was pre-sented in [12]. It was deployed at the Inria Paris across twomultistory buildings. The testbed was built using a set ofnodes; each has a number of components contained in anOtBox. These include Raspberry Pi, multiple OpenMotedevices, and a screen. Each OtBox has a testbed control soft-ware running on the Raspberry Pi and connecting to anMQTT broker using WiFi connectivity. Users can use APIsfor mote configuration and OtBox management. They alsocan use the screen to interact with the system using a web-based user interface with a dashboard running as a serviceon IBM Cloud. The testbed is an open-source and open-hardware system which provides open and remote access.

In [13], the MakeSense testbed which enables real-lifeand large-scale IoT experimentation for social research wasintroduced. It enables real-time indoor activities monitoringand situation-aware applications. The testbed design is basedon a scalable and flexible client-server model. A set of sensornodes were deployed as clients and connected to a serverusing the Lightweight Machine-to-Machine protocolthrough Broadband Internet. The server is hosted in thecloud for better IoT data analysis and visualization. Thetestbed provides API access to sensor nodes for remote

3Wireless Communications and Mobile Computing

management and configuration of available resources. Thetestbed features different security implementations includingthe Datagram Transport Layer Security (DTLS), presharedkey authentication, and data encryption using AdvancedEncryption Standard (AES).

An online testbed for large-scale LoRa experimentationsdeployed across multiple apartments and buildings was pre-sented in [14]. It has a three-layer design architecture. Theseare the node, gateway, and cloud server layers. At the nodelayer, a set of Raspberry Pi were implemented and poweredusing power banks and solar energy. The gateway node atthe next layer is responsible for packet forwarding from thenodes to the server. The server layer provides different func-tionalities including user management, system control, andjob schedule. The testbed provides web-based user interfaceenabling users to configure the nodes and receive real-timeexperimentation display.

In [15], an IoT testbed targeting effective combination ofIoT technologies and web technologies was presented. It wasdeployed in a home environment for monitoring and con-trolling different physical parameters. It has a three-layerarchitectural design which includes the sensing, middleware,and application layers. A set of sensor nodes were imple-mented using Arduino devices at the sensing layer. Thesenodes collect and send IoT data to a gateway PC (at the mid-dleware layer) using the Zigbee protocol. At the applicationlayer, a web server was implemented to enable user accessand system configuration using HTTP.

Table 1 summarizes and compares these testbed systemswith the proposed one. It can be noticed that most of the pre-sented IoT testbeds target specific IoT communication tech-nologies, sensors, and applications. This work factors in thediversity of IoT technology and the variety in IoT setupsand properties. The focus is on facilitating general-purposeand multidimensional support of IoT experimentation whileconsidering different IoT applications. It also envisages theimportance of developing an IoT testbed platform foradvancing research activities in the College of Computer atQassim University. The main focus is to build a system thatwill provide researchers with an efficient IoT facility to pro-mote more effective IoT experimentation.

3. IoT Overview

The IoT technology enables merging the physical and virtualworlds by interconnecting most of the physical objectsaround us to the Internet. It extends computing capabilityand network connectivity to everyday objects and enablingthem to generate and disseminate data. These objects couldinclude home appliances, wearable devices, medical equip-ment, and vehicles.

IoT systems are based on the combination of differenttechnological domains such as sensor technologies, network-ing protocols, embedded systems, cloud integration, dataprocessing, and communication technologies. The IoT tech-nology makes a great push towards digital intelligence andsmart control and automation with less human intervention.

The IoT technology opens the horizon for a wide scope ofsmart applications in different domains such as transporta-

tion, healthcare, agriculture, utilities, and surveillance. Smartcity is an example of IoT application that can be applied torealize better management of different aspects such as streetlight, traffic, and daily services. Another example applicationis fleet management which enables real-time tracking of fleetassets. In addition, IoT can also be applied for healthcareapplications to develop remote patient monitoring and alert-ing systems.

Most of the IoT applications require large deployment ofinterconnected IoT nodes as presented in Figure 1. Thesenodes are usually implemented with small-sized devices lim-ited in computation, memory, and energy resources. Wirelesssensor networks are typically established among these nodesto be then connected to the Internet using IP-enabled gate-ways. Each node is capable of sensing and collecting IoT datatargeting specific application. Data communications can beestablished between the IoT objects and the Internet as wellas among the objects themselves.

To this end, IoT relies on different essential hardware andsoftware components. These include sensor devices whichare important for reading varying parameters of the physicalworld. These could be environmental, meteorological, eco-logical, and medical parameters. Each node also must havea processing unit which can be built using a microcontroller/-microprocessor and memory to support data processing andstoring. The communication module is also another essentialcomponent for any IoT system to enable data transmission tothe Internet. It is evident that wireless and low power com-munication is critically required in this context as most ofthe IoT nodes would be battery-powered. Example technolo-gies for wireless and low power connectivity are Zigbee, Blue-tooth Low Energy (BLE), 6LoWPAN, and LoRaWAN. Allthese node components need to be powered using a powersupply unit which could be a battery or a solar cell.

In IoT systems, sensor nodes collaboratively sense certainparameters and communicate the readings with the gatewaywhich then forwards the data to the Internet. During this, thegateway can also be used to carry out local IoT data preprocess-ing at the edge of the IoT network. IoT data is typically sent tocloud systems in the Internet for effective data storing andman-agement. IoT applications generate big data that requires fur-ther application-specific data processing and analysis.

However, IoT poses different challenges which need to beaddressed for different IoT solutions. These include effectivesupport of security and high level of protection against potentialIoT attacks. Privacy and the ownership of collected data alsobecome critical concerns, in addition to the need for establish-ing clear privacy protection policies. Another challenge is con-cerned with legalization and having a global IoT governancesystem. It is also challenging to maintain interpretability andsupport of homogeneity in IoT systems. IoT solutions alsorequire high level of Internet availability with effective costmanagement particularly for large-scale IoT deployment.

4. System Requirement Analysis

The IoT is an interdisciplinary concept that incorporatesvariety of computing disciplines such as networking, com-munication, embedded systems, cloud computing, and data


Table1:Com

parisonof

theprop

osed

testbedwithothersystem

s.

Ref.

Scale

Objective

IoT

commun

ication

Implem

entedIoT

devices

Noof

nodes

Sensors

Interface

Cloud

integration

Data

visualization

Nod

emobility

Portability

[4]

Small

Supp

ortofIoT

applications

relatedto

environm

ental

mon

itoring

Zigbee

RPi,Arduino

,XBee

S2B

6Tem

perature,

humidity

Web

API

Not

supp

orted

Not

supp

orted

Not

supp

orted

Supp

orted

[5]

Large

Supp

ortof

multichannel

WSN

s

Dual-radioat

2.4GHz&

900MHz

RPi,op

almote,debu

gboard

100

—Web

Not

supp

orted

Not

supp

orted

Not

supp

orted

Not

supp

orted

[6]

Large

Real-lifeuser-centric

experimentalIoT

research

facility

IEEE802.15.4

TelosBmote,sensor

board,

GuruP

lug,

And

roid

phon

es,

And

roid

tablets

200

Tem

perature,

light,n

oise

level,

motion

Web

API

Supp

orted

Supp

orted

Not

supp

orted

Not

supp

orted

[8]

Small

Supp

ortof

acustom

izable

andmultipu

rpose

experimentation

2.4GHzradio

Arduino

,ATtiny85,

ATmega328P,

ESP

8266,n

RF24L

01+

——

Web

Not

supp

orted

Supp

orted

Not

supp

orted

Supp

orted

[9]

Small

Supp

ortof

flexibledata

collectionandprocessing

WiFi

RPi

10Tem

perature,

light,sou

nd,

motion

Web

API

Not

supp

orted

Supp

orted

Not

supp

orted

Supp

orted

[10]

Small

Supp

ortof

effective

evaluation

ofIoTprotocols

andalgorithms

Adho

cRPi,NotePCs

8—

API

Not

supp

orted

Not

supp

orted

Supp

orted

Supp

orted

[11]

Large

Supp

ortof

the

depend

ability

oflow-

power

WSN

sin

environm

entswithhigh

radiointerference

—RPi,TelosBmote,

NavSpark-GL,

TI

LMP92064

45Light

Web

Not

supp

orted

Supp

orted

Not

supp

orted

Supp

orted

[12]

Large

Supp

ortof

cost-effective

large-scaleIoT

experimentation

IEEE802.15.4

RPi,OpenM

ote,

displayscreen

80Tem

perature,

humidity

Web

API

Supp

orted

Supp

orted

Not

supp

orted

Not

supp

orted

[13]

Large

Supp

ortof

real-life

experimentation

forsocial

research

onindo

oractivities

BLE

ArchPro

100

Tem

perature,

humidity,light,

noiselevel,

motion,

dust

density

Web

API

Supp

orted

Supp

orted

Not

supp

orted

Supp

orted

[14]

Large

Supp

ortof

cost-effective

LoRaexperimentation

ina

large-scaleandrealistic

environm

ent

LoRaW

AN

RPi,Lo

Ramod

ule

50—

Web

Supp

orted

Not

supp

orted

Not

supp

orted

Not

supp

orted


Table1:Con

tinu

ed.

Ref.

Scale

Objective

IoT

commun

ication

Implem

entedIoT

devices

Noof

nodes

Sensors

Interface

Cloud

integration

Data

visualization

Nod

emobility

Portability

[15]

Small

Supp

ortof

effective

combination

ofIoT

techno

logies

andweb

techno

logies

Zigbee

Arduino

,XBee

—Tem

perature,

humidity,light,

gas

API

Not

supp

orted

Not

supp

orted

Not

supp

orted

Supp

orted

Propo

sed

testbed

Small-

large

General-purpo

seand

multidimension

alsupp

ort

ofIoTexperimentation

BLE

Zigbee

6LoW

PAN

LoRaW

AN

RPi,Arduino

,GrovePi,

XBee,6Lo

WPAN

mod

ule,Lo

Ramod

ule,

UPSHATbattery

12(extendable)

Awiderangeof

GrovePisensors

Web

API

Supp

orted

Supp

orted

Supp

orted

Supp

orted


science. IoT systems have distinct networking requirementsin terms of low power communication and lightweight com-puting. It also relies on easy-to-setup IoT network, cloudintegration, and data processing. This makes IoT systemchallenging to be developed and implemented. For effectivelyexperimenting the different aspects of IoT technology, thedesign of the IoT testbed should consider different character-istics. These would include efficiency, practicality, simplicity,usability, and interactivity.

IoT architecture consists of different layers as presentedin Figure 2. These are the sensing, network, data, and appli-cation layers. The testbed should be designed consideringthis architectural model in a modular design. At the sensinglayer, it is important to have a variety of sensor devices thatcan capture wide range of IoT data. For the network layer,it is required to incorporate the different IoT communicationtechnologies including Zigbee, BLE, 6LoWPAN, and LoRa-WAN. In addition, support of different networking aspectssuch as data type, network topology, node mobility, networkdensity, and node distribution should be considered. At thedata layer, the system should be integrated with a cloud sys-tem that enables effective management and processing of theIoT data. The design also needs to facilitate easy use of thetestbed to establish certain IoT applications.

Other key requirements include easy control of the sys-tem operations. Users should be able to easily configure thetestbed for different IoT scenarios and applications. The sys-tem should enable the users to have control on which sensorsto be used, communication technology to run, and applica-tion to build. It is also important to allow the users to config-ure each setup in terms of different aspects such as sensorreading interval and experiment duration. It is also impor-tant to enable the experimentation of different IoT network-ing problems such as load balancing, failure recovery, QoS,mobility, and security.

This also includes defining the scenario which each nodeshould apply. For example, a setup can be configured to emu-late a simple smart home application for 30 minutes using 5IoT nodes interconnected over a Zigbee network. Each node

is configured to read data from a temperature, pressure, light,and motion sensors every 10 seconds and send the data to thecloud. An experimental network scenario can be applied tothis setup to have a multihop network topology. During theexperiment, node failure can be triggered for certain periodin order to examine network performance.

Moreover, simple accessibility of the testbed needs to berealized for the different kinds of potential users. It is alsoimportant in this context to allow remote and online accessto the testbed. On the other hand, secure design of the systemshould be realized to at least prevent unauthenticated accessand unauthorized use of the system.

There are a number of nonfunctional requirements thatneed to be fulfilled by the system. These mainly include scal-ability, portability, flexibility, and cost-effectiveness. The sys-tem needs to be easily scaled up at the hardware and softwarelevels. Making the system effectively portable and easilydeployable is another important consideration. Differentusers can have different requirements, thus the design needsto be maintained general and flexible. It is important to keepthe development cost to the minimum without compromis-ing its functionality and quality.

5. System Design

The design of the proposed IoT testbed system is based on amodular and scalable architecture, and overview description

InternetGW

IoT network

GW

IoT network

GW

IoT network

Cloud

GW

Figure 1: Typical IoT deployment.

Sensing layer

Network layer

Data layer

Application layer

Figure 2: The layered IoT architecture.


is provided in Subsection 5.1. A set of hardware and softwarecomponents were incorporated in the system design as pre-sented in Subsection 5.2. In Subsection 5.3, a detailed opera-tional overview of the testbed functionality is presented.

5.1. Functional Design Architecture. Figure 3 presents anoverview of the testbed design architecture. The modulardesign of the testbed is realized with different functionalmodules incorporated over the different architectural entitiesof the testbed. These entities include the central server andtestbed IoT nodes. The current design of the testbed is com-posed of six main modules structured in a systematic manneras follows.

5.1.1. Access Module. The access module is the main front-end layer of the testbed. It facilitates simple and remoteaccess to the system through a web-based Graphical UserInterface (Web-GUI). It also enables user access using prede-fined Application Programming Interface (API) calls. In bothcases, the module manages secure access to the system withreliable user authentication and authorization.

5.1.2. Control Module. The control module provides centralcontrol and management for the testbed system. It is respon-sible for monitoring the health and availability of the testbedresources including the IoT nodes. It includes a node registryunit for real-time management of node state. The module isalso responsible for managing system functionality and con-trolling user experiments. It receives user requests from theaccess module once submitted through the web-GUI. Userrequest can also be received from the user directly. In bothcases, the commands are sent as API calls. Then, the controlmodule manages experiment configurations, initiates experi-ments with the nodes, and controls experimentation setups.Logs on current system state in addition to the data of exper-iment input and output are collected by the module to bethen sent to the data module for permanent storage.

5.1.3. Node Module. The node module manages the differentoperations of the node entity. It applies requested experimentconfigurations as received from the control module ordirectly from the user using the system’s APIs. The nodemodule is then responsible for running and controllingexperimentation setups. It also communicates real-time dataof experiment output with the control and data modules.

5.1.4. ExperimentationModule. The experimentation modulecontains the sensing, communication, and energy units of thenode entity. These are the main elements for setting up IoTexperiments and applications on the testbed. The experimen-tation module controls the operations of these units accord-ing to the commands received from the node module. Oncerunning an experiment, the module collects sensor readings,communication measures, and energy consumption infor-mation. These outputs are then delivered to the node modulein a real-time manner.

5.1.5. Data Module. The data module maintains the requiredfunctionality for data storage and management. It is respon-sible of permanently storing system logs, experiment input,and collected experimentation data in the database unit. Italso manages user data for enabling secure access to the sys-tem. The data module also provides effective integration withweb-based cloud systems. The output data of user experi-ments can be forwarded to the cloud for further data man-agement and processing.

5.1.6. Network Module. The network module provides theunderlying communication infrastructure to interconnectmost of the system modules. It enables wired and wirelessconnectivity to the system via WiFi and Ethernet, respec-tively. The network module is responsible for establishing asystem control network among the testbed entities, users,and integrated cloud systems.

5.2. Main Design Components. The testbed design architec-ture is composed of two main architectural entities of the

Network moduleUser

Web-GUI

API

Registry

APIAPI

WiFi Ethernet

Data module

DB Cloud

Sensing

Comm.

Energy

Exp.module

Nodemodule

Controlmodule

Accessmodule

Testbed nodesTestbed server

Figure 3: Overview of the testbed design architecture.


testbed. These are the central server and testbed IoT nodes.Both entities are interconnected over a wireless control net-work providing permanent Internet access.

The server entity consists of a set of functional compo-nents as shown in Figure 4. These include the system control-ler, web server, and database:

(i) The controller runs the functionality of the controlmodule at the server entity. It facilitates effectivetestbed management and experimentation controlat the main back-end layer of the system. It main-tains a node registry service for managing the avail-ability and monitoring the performance of thetestbed nodes

(ii) The web server provides the functionality of theaccess module at the server entity. It enables web-based access to the testbed system in addition tofacilitating easy configuration and setup of the dif-ferent testbed resources. It also provides the suffi-cient support to prevent unauthenticated andunauthorized user access

(iii) The server manages a database that provides localdata storage as part of the data module. It mainlyretains user access information, system logs, andexperiment data

The testbed node entity represents a set of IoT nodedevices. Each of these nodes has a number of hardware andsoftware components as shown in Figure 5. The hardwarecomponents are the sensors, communication devices, andpower source:

(i) The sensors include a set of IoT sensor device thatmeasures a set of parameters such as environmental,biochemical, physiological, behavioral, and visualones

(ii) The communication devices enable establishingwireless IoT networks among the IoT nodes. Themain IoT communication technologies currently

supported by the testbed are LoRaWAN, 6LoW-PAN, Zigbee, and BLE

(iii) The design of testbed also relies on portable and sus-tainable power unit that allows for monitoringpower consumption and controlling remainingenergy

These components of the IoT nodes support easy recon-figuration, lightweight portability, cost-effective implementa-tion, and effortless deployment. The experimentationmodule is completely run at the node entity to manage theoperations of the hardware components. It enables eachcomponent to apply relevant user configurations for effec-tively running the requested experimental setup.

The node entity also includes a central software compo-nent which is the node manager. It runs the functionality ofthe node module to control the node operations. These oper-ations include initial node registration and frequent statusreporting with the system controller. In addition, the nodemanager controls user experiments according to therequested setup and component configurations. For example,it manages turning on/off specific communication module,initiating sensor reading, and communicating IoT data. Thenode manager also performs real-time collection and trans-mission of experiment data to the system controller at theserver. The data can also be sent by the node manager tothe cloud if required by the user.

For better system usability, the design facilitates easy andsimple access via an interactive web-GUI, which enables thefollowing:

(i) Resource discovery and reservation: the systemallows the reservation of a certain number of thenodes, that are currently available, for each experi-mentation setup

(ii) Experiment configuration and execution: users canselect which communication modules and sensordevices to run for each selected node. There are alsoentries for setting experiment duration and sensinginterval. Users can also provide experimental scriptto apply certain experiment scenario on the selectedresources. This enables setting network topology,setup mode, data traffic type, and packet size

(iii) Topology display: the interface incorporates anexperiment representational display for real-time

DB

Noderegistry

WiF

iEt

hern

et

Webserver

Testbed server

APIserver/client

System controller

Web-GUI

Figure 4: The architectural components of the server entity.

LoRa

WA

N

Sensors

Zig

beeW

iFi

Ethe

rnet BL

E

Node manager

Testbed nodes

Powerunit

6LoW

PAN

API

serv

er/c

lient

Figure 5: The architectural components of the IoT node entity.


visualization of the currently running setup. It visu-alizes the flow of IoT data among the experimenta-tion resources during each run

(iv) Data dashboard: the interface provides visual displayof collected IoT experiment data in a real-timemanner

The system also includes other software componentssuch as the API servers and clients at the server and nodeentities. User requests and experiment configurations areexchanged among the server and client using predefinedAPI calls. After the initial startup time, each node runs anAPI server that waits for experimentation calls from theAPI client. User input via the web-GUI is firstly parsed intoAPI calls and then transmitted by the API client to the APIserver. These are then processed and forwarded to the nodemanager for running the requested experimentation setup.The API server also accepts API calls from users directly inthe case of having no functioning control module.

The communication of API calls among the differententities is maintained using the network module. This is car-ried out over a wireless control network established using aprivate WiFi access point. Using wired Ethernet connectivityis also supported as an alternative option. In both cases, theentities are provided with Internet access.

5.3. Operational Overview of the System. The sequence dia-gram presented in Figure 6 provides an overview of the mainoperations of the testbed system. At the startup stage, eachnode discovers its currently available experimentation mod-ules and hardware components. At the node entity, the nodemodule sends a status check message to the experimentationmodule which responds with status information of the avail-able experimentation resources. Then, the node modulesends a registration request message to the control module.This message contains different node information includingits ID, IP address, available experimentation components,and remaining power. The message is processed by the noderegistry unit to maintain the availability and monitor the per-formance of the testbed nodes. It then replies to the nodemodule with a response message. The node information isalso sent to the data module for permanent storage of thedata in the database.

User access to the system is managed by the access mod-ule during the login stage. New users are required to firstlyregister to the system. The registration information is for-warded to the data module for permanent storage. Afterreceiving access request, user data is obtained from the localdatabase to carry out secure user authentication and authori-zation. Users receive access tokens if and only if the access islegitimate.

User requests for IoT experimentations can be issuedthrough the web interface or API calls. In the former case,experiment configurations and setup script entered by theuser are forwarded to the control module in a request mes-sage. After copying the request data into the database, thecontrol module forwards the message to the node moduleusing internal API calls. The request is processed by the node

manager software at the testbed nodes selected for the cur-rent experimentation. It then replies with a response messagewhich is forwarded by the control module to the display unitfor user notification. Upon that, the node manager sends theuser configurations to the experimentation module to config-ure the sensing, communication, and energy componentsaccordingly. After that, a confirmation message is sent backto the node module in order to signal the readiness of theexperimentation resources. The node manager can then issuea run command and get the experiment started. It alsoapplies the setup script to control the experimentation sce-nario accordingly.

For the experiment duration, the experimentation mod-ule keeps sending the collected IoT data and measures tothe node module in a real-time manner. These are then for-warded to the control module which forwards them to theaccess module for real-time visualized display on the webinterface. The result data is also sent to the data module forlocal storage and cloud backups.

In the other case, users issue experimentation requestsdirectly to the testbed nodes using API calls. This supportssystem availability during the failure of the server entity.The request message is received and processed by the nodemodule at each of the selected IoT nodes. The node managerissues a response to the user, copies the request data to thedata module for cloud storage, and then sends the user con-figurations to the experimentation module. It also initiatesthe experiment after receiving a setup confirmation messagefrom the experimentation module. Once the experimentstarted, the collected IoT experimentation data is forwardedto the node manager in a real-time manner. The data is thenforwarded to the data module to be uploaded to the cloudsystem for further data processing and visualization. The userthen can freely access the data at the cloud anytime the datamodule using the given cloud API calls.

6. System Implementation

The proposed IoT testbed system was implemented based onintegrating a set of hardware components and developingdifferent software modules. A representational overview ofthe testbed implementation is presented in Figure 7. The fol-lowing subsections, 6.1 and 6.2, provide a thorough descrip-tion of the current hardware and software implementations,respectively.

6.1. Hardware Implementation Overview. A collection ofopen-source and cost-effective hardware components wasutilized to develop the current implementation of the testbed.These include a powerful PC laptop which was temporarilyset up as the testbed server. It is wirelessly connected to a pri-vate WiFi Access Point (AP) which provides access to theInternet and the cloud service. The IoT nodes of the testbedwere implemented using multiple Raspberry Pi boards ofVersion 3 and Model B. The nodes are wirelessly connectedto the AP to establish a control network with the server andgain Internet access.

Each Raspberry Pi board runs a stable and open-source Linux-like operating system, namely, Raspbian.


The decision of adopting the Raspberry Pi board wasmade considering its capabilities in terms of CPU, mem-ory, and connectivity. It also has a built-in WiFi andBluetooth Low Energy (BLE) modules which are essentialcomponents of the testbed. In addition, it supports easyintegration of wide range of IoT sensors and communica-tion modules over its I/O interfaces such as the General-Purpose Input/Output (GPIO) pins and serial ports. Thesewere used in the current implementation to attach the fol-

lowing external hardware modules to each Raspberry Piboard:

6.1.1. XBee S2C. XBee S2C is an RF module developed byDigi International to support short-range wireless Zigbeecommunication.

6.1.2. 6LoWPAN Chip. A tiny 802.15.4 radio chip was devel-oped by OpenLabs to provide support for 6LoWPAN over

User

Registration requestNode information

Registration response

Status check msg

Status information

Login requestUser data request

Access tokenUser information

Experiment configurationExperiment request

Experiment request Experiment data

Devices configurations

Setup confirmation

Run command

Experiment resultsExperiment results

Experiment results

Experiment results

Node response Node response

Visual IoT data

Experiment request

Node response

Devices configurations

Setup confirmation

Run command

Experiment results

Experiment data

Experiment results

Results request Expe

rimen

tatio

n us

ing

API

Expe

rimen

tatio

n vi

a the

web

inte

rface

Use

r acc

ess

Star

tup

stage

Results data

Accessmodule

Datamodule

Controlmodule

Nodemodule

Exp.module

Figure 6: Operational overview of the testbed system.


802.15.4 to the Raspberry Pi board. This RF module isdesigned to be attached to the GPIO pins of the board.

6.1.3. Arduino Board. Arduino is a microcontroller boardthat provides an open-source development environment. Ithas a number of digital and analog I/O pins which can beused for further extensions such as sensor devices and com-munication modules. There are different versions of Ardu-ino, and the Arduino UNO version is the one being used inthe current implementation.

6.1.4. LoRaWAN Arduino Shield. The LoRaWAN shield wasdeveloped by Dragino for Arduino UNO boards. It providesa long range transceiver that is based on open-source library.The shield has a set of I/O pins which were used for directattachment to the Arduino board.

6.1.5. GrovePi+. GrovePi+ is an add-on sensor board devel-oped by Dexter Industries to provide a simple solution forsensor attachment. It provides 7 digital, 3 analog, and 3 I2Cports. The testbed has GrovePi+ boards mounted on eachRaspberry Pi board.

6.1.6. Collection of Compatible Sensor Devices. These are con-nected to the different ports of the GrovePi+ board. Theseinclude temperature, humidity, barometer, dust, air quality,gas, light, sound, and motion sensors. However, users areprovided with the flexibility to replace these sensors with avariety of alternative sensors to support different IoTapplications.

6.1.7. External Power Module. The testbed provides two dif-ferent powering options for the battery modules. These area portable power bank and the UPS HAT Battery AdapterPower Supply Extension Board. The testbed also supportsthe integration of any other power sources such as integratedsolar cells.

Each Raspberry Pi board in the current implementationhas a GrovePi+ board attached using the GPIO pins. Thesensor devices are directly connected to the ports of the Gro-vePi+ board. The 6LoWPAN module is also attached to theextended GPIO pins on the GrovePi+ board. On the otherhand, The XBee S2Cmodule is attached to a customized don-gle which is connected to the Raspberry Pi board via its serialport. The communication between the board and the chip isestablished over a serial connection.

In addition, another serial port is used to connect eachRaspberry Pi board to an Arduino board which is used tomount and operate the LoRaWAN module. This can alsobe used as a further development board to support thescalability and flexibility of the system. However, theArduino board and LoRaWAN module are completelycontrolled by the experimentation module implementedin the Raspberry Pi. This is maintained over a permanentserial connection.

In regard to the power module, it can be a power bankthat is connected to the micro-USB power input of the Rasp-berry Pi board. It also could be the Battery Adapter PowerSupply Extension Board which can be attached to the GPIOpins.

Power bank

GrovePi board

Gro

vePi

sens

ors

RPi (Raspbian)

ThingsBoard

Cloud

LoRaWANshield

Arduino

LoRaWAN control software

Built-in BLEBuilt-in WiFi

Serial connection

GPIO

USB dongle

Power port

I/O pins

Serial interface

Serial interface

DB

(MongoDB)

6LoW

PAN

WiFiAPBuilt-in WiFi

APImanager(Node JS)

Systemcontroller(Node JS)

Web-GUI(HTML, CSS, JS)

PC (server)

APImanager(Flask-

RESTful)

Nodecontrol

(Python)

Exp.

cont

rol

XBee(Zigbee)

Figure 7: Representation of the current implementation of the system.


All the hardware components of each IoT node are con-tained in a compact-sized and lightweight box in order tosupport the portability of the system. Users are provided withthe ability of placing and moving the IoT nodes in a flexiblemanner. The current implantation was developed with 12IoT nodes. However, the scalable design and flexible imple-mentation of the system allows for easily integrating addi-tional IoT nodes in future.

6.2. Software Implementation Overview. A functional imple-mentation of the main modules of the testbed design wasdeveloped over the different hardware components of thesystem. This was accomplished using a set of software devel-opment tools as follows:

6.2.1. Access Module. The web-GUI was developed as thefront-end of the system using HTML, CSS, and JavaScript.

6.2.2. Data Module. The database was implemented usingMongoDB. For the cloud service, the current implementationis integrated with ThingsBoard, a public cloud platform. Itprovides a set of APIs which was used for managing the com-munications with the cloud service.

6.2.3. Control Module. The control module’s functionalitieswere implemented as the system back-end using Node JS.

6.2.4. Node Module. The different functionalities of the nodemodule were implemented using Python and Flask-RESTful.

6.2.5. Experimentation Module. The communication, sensor,and battery submodules were also developed using open-source Python classes. As it was attached to an Arduinoboard, however, the development of the LoRaWAN submo-dules was carried out in the Arduino C-like environment.

The current implementation of the testbed serverincludes a web-GUI run by the access module. The interfaceenables a testbed user to input a setup script that describesthe main experimentation setup of an experiment. The scriptis formed according to the user selections of the number ofnodes, communication module, sensors, setup mode, sce-nario type, and experiment duration. A variety of experimentconfigurations can also be entered to configure the compo-nents of the established setup. These include sensing interval,transmission interval, packet size, battery level threshold,node operational mode, communication channel, and net-work topology.

The testbed server also incorporates a Node JS imple-mentation of the system controller which provides the differ-ent functionalities of the control module. It runs the noderegistry which relies on a developed heartbeat mechanismfor effective remote monitoring of the IoT node health andavailability. The mechanism was implemented to frequentlyexamine the server node connectivity to detect any failure.

On the other hand, the system controller was also imple-mented to receive setup script and experiment configurationsin a POST request. The request data is then copied into theexperiment table of the system database and forwarded tothe selected nodes. The control module handles the commu-nications with the IoT nodes using a set of developed REST

API calls. Received API requests by the node module are han-dled by the Python implementation of the node control soft-ware. It processes the setup script to control the experimentaccordingly. The experiment configurations are obtained bythe experimentation control module as input to configureits components accordingly. Once the experiment is run, dataare obtained and processed by the node control software asJSON data. It is then sent back to the server in response tothe API request. The data is frequently obtained by the accessmodule for real-time display on the interface.

7. Use Case

The current implementation of the proposed IoT testbed sys-tem enables running effective experiments to examine differ-ent aspects of the IoT technology. These include thecharacteristics of IoT communication in regard to differentparameters such as link quality, transmission range, powerconsumption, and data rate. This can be carried out in exper-imental scenarios defined according to different consider-ations including setup size, network topology, hardwareconfigurations, and traffic type. Users can select to run exper-imental setups using one of the implemented IoT communi-cation technologies, namely, Zigbee, 6LoWPAN, LoRaWAN,and BLE. These would also facilitate studying different prop-erties such interference and noise for different IoT networksetups. In addition, the portability of the testbed nodesenables customized node positioning and allows for flexibleestablishment of mobility setups and dynamic scenarios. Itfacilitates studying network performance during the mobilityof certain nodes or considering the movement of the wholenetwork under realistic conditions.

Experiment can also be carried out for effective IoT datacollection and processing. Since the testbed provides manysensing options using different sensor devices, data can beeasily acquired for environmental, meteorological, ecological,and many other purposes. The integration with the cloudsystems enables effective processing and simple visualizationof the collected data. Therefore, a variety of IoT applicationscan be established using the testbed for different purposessuch as monitoring, control, or automation. Examplesinclude indoor environmental quality measurement, humanactivity monitoring, smart campus, indoor asset tracking,and smart classroom management.

To present a usage example of the testbed, an experimentwas carried out to measure energy consumption of selectedIoT communication technologies. In this experiment, thetestbed was accessed using API calls which are received andprocessed by the nodes directly. Output data was thenobtained from the node after having the experiment com-pleted. Table 2 lists the main setup and configurations ofthe experiment. It was carried out over a simple networktopology of only five IoT nodes. Three networking setupswere developed to interconnect the nodes in a start topologyusing different IoT communication technologies. These areBLE, Zigbee, and 6LoWPAN. In all the scenarios, however,each node was configured to sense and send temperature sen-sor data every 5 seconds during a period of 45 minutes. Eachsetup was repeated for 10 times, and the battery of each node


was fully recharged for each run. Data indicating the con-sumed energy is collected for each run, and the average con-sumption for each setup is presented in Figure 8. It can benoticed that low energy consumption was maintained in allthe setups. However, data transmission over BLE relativelyconsumed less battery power compared to the Zigbee and6LoWPAN setups.

The testbed system is also now being used for a new IoTexperiment which requires the deployment of the testbednodes across the building of the College of Computer at Qas-sim University. The main objective of the experiment is toexperimentally examine the effect of IoT node positioningon the overall network performance in indoor environments.The focus is on two different IoT communication technolo-gies which are Zigbee and LoRaWAN. Ten nodes are cur-rently utilized to run different scenarios with differenttopological setups. In each of these scenarios, the nodes areplaced in varying locations inside the building. In some ofthe scenarios, some of the nodes will be mobile with a simplemovement pattern. This would enable to establish differentsetups considering different criteria such as topological dis-tance and the presence of obstacles.

8. Evaluation

For effective evaluation, the system was assessed using theTechnology Acceptance Model (TAM) [16]. It is a com-monly adopted method for analyzing technology usagebehavior and user attitude. Accordingly, a TAM-based ques-tionnaire was developed for this study in respect to the com-mon TAM factors. The questionnaire was then filled by atotal of 46 participants who were involved in the use of thesystem.

The TAM-based questionnaire was formed with the mainTAM factors. These are as follows: perceived usefulness, per-ceived ease of use, attitude toward usage, and behavioralintention to use. The questionnaire basically measures 16items according to a 5-point rating scale ranging from “1:strongly disagree” to “5: strongly agree”. Figure 9 presentsthe TAM model of the system.

The participants were grouped in different groups of 2-3participants, and some of the participants used the system

individually. There were a total of 16 different groups and 4individuals. There was good diversity among the participantsin terms of age, educational level, and technical knowledge.That is, age range of the participants was 19-48 years. About70% of the participants were MSc and BSc students, whereasthe rest were PhD and MSc degree holders. In terms of theirtechnical knowledge, most of the participants had no IoTexperience, whereas few of them had some IoT knowledge.

The proposed testbed was used by all the participants totest the different functionalities of the testbed. This took placein one of the labs at the College of Computer in Qassim Uni-versity. The participants were involved in building and run-ning three IoT experiments considering IoT data collectionand processing, IoT networking using the different commu-nication technologies, and the development of IoT applica-tions. The main details of these experiments are presentedin Table 3. Each experiment was developed for certain userlevel, ranging from basic to advanced, and configured forspecific testing duration. The experiments also varied interms of the number of nodes, communication technology,and topology setup. The participants also used different typesof sensors with varying sensing interval in each experiment.They also experienced different modes to access and config-ure the system. It is important to note that every participantgroup and individual was involved in at least two of theexperiments; each was conducted for only a single run.

The evaluation was based on the observation of the par-ticipants during their use of the testbed. The TAM question-naire was also filled by every participant individuallyafterwards. After collecting the evaluation data, the reliabilityof the TAM results was firstly evaluated. This was based oncalculating Cronbach’s alpha coefficient for each of the con-sidered factors. It is a consistency coefficient representingthe dimensionality of the TAM questionnaire’s questions.The resulted coefficient values are listed in Table 4. As itcan be seen, high value of Cronbach’s alpha coefficient wasachieved by each factor. There was no coefficient value lowerthan the minimum acceptable value of 0.7.

On the other hand, Table 5 shows the evaluation resultsof the TAM questionnaire. It lists the resulted values of thecalculation of the scoring average and standard deviationconsidering each TAM factor. It can be seen that the pro-posed IoT testbed was able to achieve good evaluation results

Table 2: Experiment configurations.

Parameter Setting

Number of nodes 5 nodes

Communication modules BLE, Zigbee, and 6LoWPAN

Scenario Data transmission

Duration 60 minutes

Transmission interval 4 seconds

IoT data type Temperature sensor readings

Data packet size Low-sized payload

Setup mode Static

Battery charge status Fully

Network topology Star

Number of runs 10 runs

BLE Zigbee

Communication technology scenario

6LoWPAN0

5

10

15

20

Con

sum

ed b

atte

ry p

ower

(%) 25

Figure 8: Experimental results.


as high average scores and low standard deviation wasobtained for all the factors. It is evident that the testbed suc-ceeded in maintaining usefulness and simplicity when itcomes to realistic IoT experimentation. In addition, thetestbed provides an attracting and motivating experimentalIoT facility for the different users as they showed a positiveattitude toward the testbed with good intention to use it infuture for IoT experimentation.

Furthermore, the correlation between the consideredTAM factors was measured and analyzed based on the calcu-lation of the Pearson correlation (PC). This is a critical statis-tical indicator for analyzing how the dependency relationship

formed among the factors. The calculation results of thePearson correlation are listed in Table 6. It can be noticedthat there was a positive correlation among the consideredTAM factors for the proposed IoT testbed. The perceivedusefulness factor well-correlated with the ease of use factor.Both of these factors had good correlation with the attitudetoward the usage of the system. It is accordingly evident thatthe testbed attracted the different users to utilize its function-ality as a result of its simplicity and efficiency for effective IoTexperimentation. In addition, the relationship among theattitude toward usage and intention to use factors achieveda higher correlation value. This clearly indicates that the

Perceivedusefulness of the

IoT testbed

Perceived ease ofuse of the IoT

testbed

Behavioralintention to usethe IoT testbed

Actual use of the IoT testbed

Attitude towardsusing the IoT

testbed

Figure 9: The TAM model of the proposed IoT testbed.

Table 3: The main details of the experiments for the TAM evaluation.

Setting Exp. 1 Exp. 2 Exp. 3

Experiment type Smart home application IoT data processing IoT networking

Level Basic Intermediate Advanced

Duration (min) 25 35 45

No. of nodes 6 4 8

Communication tech. BLE ZigbeeZigbee

6LoWPANLoRaWAN

Network topology Star StarStar

Multihop

SensorsLightMotionGas

Temperature humidity air quality Temperature

Sensing interval (sec) 5 10 5

Data packet size Small Small Small-large

Access mode Web-GUIWeb-GUI

APIAPI

Data storage CloudCloudLocal

Local

Participant no.13 groups

3 individuals15 groups

4 individuals8 groups

1 individual

Table 4: Cronbach’s alpha coefficient results.

Factor Cronbach’s alpha coefficient

Perceived usefulness 0.914

Perceived ease of use 0.882

Attitude toward usage 0.931

Behavioral intention to use 0.844

Table 5: Average scoring results.

Factor Average STD

Perceived usefulness 4.41 0.71

Perceived ease of use 4.34 0.76

Attitude toward usage 4.25 0.68

Behavioral intention to use 4.23 0.79


system was able to maintain high level of user satisfactionand attract user interest for future IoT experimentation.

To determine the variance of the independent factors, theR-square is also calculated and listed in Table 6. It indicatesthe closeness of the data to the fitted regression curve.Figure 10 shows the calculated R-square values on the TAMmodel of the proposed IoT testbed. It is evident that theseresults statistically present effective variation in the responsevariable.

9. Conclusion

An effective IoT testbed system that takes IoT experimenta-tion to a further level of simplicity and practicality was pre-sented in this paper. The system provides a motivatingexperimental testbed environment for effective IoT experi-mentation. It enables researchers to carry out IoT experi-ments targeting the different aspects and applications of theIoT technology. The testbed design provides a usable, scal-able, portable, and flexible system. Testing and experiment-ing IoT solutions was made inspiring in simple andpractical design architecture. The current implementationof the testbed system was accomplished using a collectionof efficient and cost-effective components. The results ofthe TAM-based evaluation demonstrated how the proposedtestbed system is usable and useful for conducting experi-mental IoT studies in a realistic manner. The testbed receivedpositive attitude in regard to its efficiency for IoT experimen-tation. The testbed was also found to be effective for futureuse in future IoT research works.

Given its design modularity and flexibility, the testbedcan be extended to target more customized IoT experiments.In a future work, the testbed will be utilized to carry out anumber of IoT experiments to study IoT network perfor-mance considering dynamic and mobile network setups. Itis also envisaged that the system can be deployed for a smart

classroom application which would require integrating andcomparing some artificial intelligence techniques.

Data Availability

According to the funding policy of this work, data cannot beshared or made publicly available during the fundingcontract.

Conflicts of Interest

The author declares that there is no conflict of interestregarding the publication of this paper.

Acknowledgments

This work was supported by the Deanship of ScientificResearch, Qassim University, according to the agreement ofthe funded project No. SRD-coc-2018-1-14-S-5142. Theauthor thanks the sponsor of this work for their support.

References

[1] G. Z. Papadopoulos, A. Gallais, G. Schreiner, E. Jou, andT. Noel, “Thorough IoT testbed characterization: fromproof-of-concept to repeatable experimentations,” ComputerNetworks, vol. 119, pp. 86–101, 2017.

[2] M. Hossain, S. Noor, Y. Karim, and R. Hasan, “IoTbed: ageneric architecture for testbed as a service for Internet ofThings-based systems,” in 2017 IEEE International Congresson Internet of Things (ICIOT), Honolulu, HI, USA, June 2017.

[3] T. W. Edgar and T. R. Rice, “Experiment as a service,” in 2017IEEE International Symposium on Technologies for HomelandSecurity (HST), Waltham, MA, USA, April 2017.

[4] S. Ferdoush and X. Li, “Wireless sensor network system designusing Raspberry Pi and Arduino for environmental

Table 6: Relationships among TAM factors.

Independent factor Dependent factor PC R2

Perceived ease of use Perceived usefulness 0.798 0.67

Perceived usefulness Attitude toward usage 0.817 0.72

Perceived ease of use Attitude toward usage 0.791 0.69

Attitude toward usage Behav. intention to use 0.836 0.70

Perceivedusefulness of the

IoT testbed

Perceived ease ofuse of the IoT

testbed

Attitude towardsusing the

IoT testbed

R2 = 0.72

R2 = 0.67

R2 = 0.69

R2 = 0.70

Behavioralintention to usethe IoT testbed

Figure 10: The TAM model with the R-square values.


monitoring applications,” Procedia Computer Science, vol. 34,pp. 103–110, 2014.

[5] Q. Li, D. Han, O. Gnawali, P. Sommer, and B. Kusy, “Twonet:large-scale wireless sensor network testbed with dual-radionodes,” in Proceedings of the 11th ACM Conference on Embed-ded Networked Sensor Systems, Roma, Italy, November 2013.

[6] M. Nati, A. Gluhak, H. Abangar, and W. Headley, “SmartCampus: a user-centric testbed for Internet of Things experi-mentation,” in 2013 16th International Symposium onWirelessPersonal Multimedia Communications (WPMC), AtlanticCity, NJ, USA, June 2013.

[7] L. Sánchez, J. A. Galache, V. Gutiérrez et al., “SmartSantander:experimentation and service provision in the smart city,” inProceedings of the Future Network & Mobile Summit, Warsaw,Poland, June 2011.

[8] A. Di Nisio, T. Di Noia, C. G. C. Carducci, andM. Spadavecchia, “Design of a low cost multipurpose wirelesssensor network,” in 2015 IEEE International Workshop onMeasurements & Networking (M&N), Coimbra, Portugal,October 2015.

[9] K. Hentschel, D. Jacob, J. Singer, and M. Chalmers, “Supersen-sors: Raspberry Pi devices for smart campus infrastructure,” in2016 IEEE 4th International Conference on Future Internet ofThings and Cloud (FiCloud), Vienna, Austria, August 2016.

[10] H. Oda, E. Kulla, R. Ozaki, and N. Nishihara, “Design of anadhoc testbed for IoT and WSAN applications using Rasp-berry Pi,” in Proceedings of the 11th International Conferenceon Broad-band Wireless Computing, Communication andApplications (BWCCA), Asan, Korea, November 2016.

[11] M. Schuß, C. A. Boano, M. Weber, and K. Römer, “A compe-tition to push the dependability of low-power wireless proto-cols to the edge,” in Proceedings of the 14th InternationalConference on Embedded Wireless Systems and Networks(EWSN), Uppsala, Sweden, February 2017.

[12] J. Munoz, F. Rincon, T. Chang et al., “OpenTestBed: poorman’s IoT testbed,” in IEEE INFOCOM 2019 - IEEE Confer-ence on Computer Communications Workshops (INFOCOMWKSHPS), Paris, France, France, April-May 2019.

[13] J. Jiang, R. Pozza, N. Gilbert, and K. Moessner, “MakeSense: anIoT testbed for social research of indoor activities,” 2019,http://arxiv.org/abs/1908.03380.

[14] W.Wang, X. Guo, L. Liu et al., “Dandelion: design of an onlinelarge scale LoRa testbed,” in Proceedings of the InternationalConference on Embedded Wireless Systems and Networks(EWSN), Beijing, China, February 2019.

[15] A. A. Reshi, A. Alsaeedi, and S. Shafi, “Development and webperformance evaluation of Internet of Things testbed,” in2019 International Conference on Computer and InformationSciences (ICCIS), Sakaka, Saudi Arabia, April 2019.

[16] F. D. Davis, “Perceived usefulness, perceived ease of use, anduser acceptance of information technology,” MIS Quarterly,vol. 13, no. 3, pp. 319–340, 1989.


http://arxiv.org/abs/1908.03380

a multidimensional internet of things testbed system

Documents