![IoT hardware compatible with TensorFlow Lite. Adapted from [6].](https://www.researchgate.net/profile/Sergio-Lopes-7/publication/350450778/figure/tbl1/AS:1027326963896322@1621945243682/IoT-hardware-compatible-with-TensorFlow-Lite-Adapted-from-6_Q320.jpg)
Content uploaded by Sérgio Ivan Lopes
Author content
All content in this area was uploaded by Sérgio Ivan Lopes on May 25, 2021
Content may be subject to copyright.
Designing a FIWARE-Based Smart
Campus with IoT Edge-Enabled
Intelligence
Pedro Martins1,SérgioI.Lopes
1,2(B
),andAntónioCurado
3
1ADiT - Instituto Politécnico de Viana do Castelo,
4900-348 Viana do Castelo, Portugal
2IT - Instituto de Telecomunicações, Campus Universitário de Santiago,
3810-193 Aveiro, Portugal
sil@estg.ipvc.pt
3Prometheus - Instituto Politécnico de Viana do Castelo,
4900-348 Viana do Castelo, Portugal
Abstract. Higher education institutions are passing through a fast dig-
ital transformation process that has the potential to enable frictionless,
touchless, and more intuitive experiences in academia. Moreover, stu-
dents are now digital natives and demand from higher education insti-
tutions new digital services for all academic purposes. In this article, we
introduce the design methodology used for the architecture specification
of the IPVC Smart & Sustainable Campus (IPVC-S2C), a FIWARE-
based platform with edge-enabled intelligence. The current research also
surveys and characterizes low-cost IoT edge hardware capable of per-
forming distributed machine learning. Lastly, a proof of concept focus
on Indoor Air Quality monitoring on the campus is presented and the
forthcoming research is outlined.
Keywords: Smart campus ·IoT ·Context-driven ·Edge-intelligence
1 Introduction
The higher education landscape is passing through a continuous and fast digi-
tal transformation process that is enabling frictionless, touchless, and intuitive
experiences in academia. Students who are nowadays digital natives push the
higher education institutions into the digital transformation and at the same
time foster new digital services that will turn into reality a fully and connected
digital campus experience. A digital campus also referred to as a smart campus,
takes advantage of existing Information and Communication Technologies (ICT)
and state-of-the-art Internet of Things (IoT) technologies to provide automated
and intelligent services on the campus.
In this article, we introduce the core elements and the ICT infrastructure
needed to support the IPVC Smart & Sustainable Campus (IPVC-S2C) imple-
mentation and put forward a FIWARE-based architecture that enables full inte-
gration with other legacy systems that higher education institutions still have
c
⃝The Author(s), under exclusive license to Springer Nature Switzerland AG 2021
Á. Rocha et al. (Eds.): WorldCIST 2021, AISC 1367, pp. 557–569, 2021.
https://doi.org/10.1007/978-3-030-72660-7_53
558 P. Martins et al.
in use, by taking a case study as an example, cf. Sect. 5. Moreover, a set of core
criteria for the selection of IoT edge hardware capable of performing distributed
machine learning is also included for future edge intelligence integration.
The remainder of this document is organized as follows: Sect. 2presents
existent frameworks/platforms used in the development of smart solutions in the
context of a smart campus; Sect. 3introduces the IPVC-S2C ICT infrastructure;
Sect. 5introduces the IPVC-S2C architecture and describes in detail its core
elements; lastly, in Sect. 7the main conclusions of this work are introduced and
future work guidelines presented.
2 Related Works
SmartCAMPUS UWB project develops the concept of a reduced scale model
of a city that enables the creation of a testbed for smart and IoT technologies.
Currently, it involves 9 faculties equipped with LoRaWAN and Sigfox communi-
cation infrastructures. Several applications are already in use inside the campus,
such as KETCube, which is a prototype and demo platform; Environment, on
which it is possible to detect fluctuations in temperature or humidity; Cloud
which collects and evaluates information from the implemented systems; Park-
ing which manages car parking with IoT sensors; IoT Lab which is a lab devoted
to students developing IoT projects and offers a Data Warehouse with open data
captured by several IoT Devices [1].
WiseTown, “Web Information Streams Enhancer for Your Town” is an appli-
cation that uses FIWARE and collects information from different data sources,
making data easy to manage and the information organized, supporting the Pub-
lic Administration by improving urban planning, modernizing public services,
and streamlining city management [2].
SmartMetropolis is a project developed in the Digital Metropole Institute
(IMD) in the Federal University of Rio Grande do Sul. The main goal of this
project is the development of methods and techniques to support the implemen-
tation of services to be integrated by smart cities by creating applications to
support strategic areas, such as security, tourism, public transportation, edu-
cation, big data, cloud computing, and IoT. IMD current projects include the
monitoring of Water and Energy, using sensors that communicate via GPRS;
Smart Place, which manages air conditioners and the lightning inside buildings,
allowing the management which reduces the average cost of the electricity bill;
SIGNatal, an open data application that shares geographical data from Natal
city with users; and finally the ROTA-Viatura, a system of dispatching police
vehicles and policemen for a faster attending on police emergency occurrences [3].
Bettair is a platform that, as described in [4], works as a service and allows
the mapping of air pollution in cities with the help of the Bettair Static Nodes, an
autonomous device that is installed in streetlights. With the information recov-
ered from the sensors, urban planners can take action to improve air quality.
Before using FIWARE, the platform was developed with a monolithic archi-
tecture, which was unable to be scaled and modified as requested by using a
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 559
FIWARE-based architecture, the Context Broker manages air pollution infor-
mation that is delivered by the sensors, accelerating the development of a Smart
Solution, offering a possibility to connect with other existing FIWARE platforms
that the city might already have.
In [5], the authors present the architecture of the University of Málaga Smart
Campus that has been designed to transform its university campuses into a small
smart city that can support efficient management of their area as well as inno-
vative educatio nal a nd research activities. The actions of t his i nitiative follow
six main application categories: 1) Emissions, Energy and Water, 2) Nature
and Environment, 3) Health and Well-Being, 4) Mobility, 5) ICTs, 6) Research,
Teach i n g , a n d I n n ova t i on. The authors prop o s e a n a r chitecture that uses IoT
technologies and several existent telecommunications resources to deliver a uni-
fied infrastructure that is used in several application domains that can be inte-
grated into learning activities.
3 IPVC Smart and Sustainable Campus
The Polytechnic Institute of Viana do Castelo (IPVC) is a higher education insti-
tution created in 1980, serving the Northwest region of Portugal. The IPVC has
six campuses spread through four places in Alto-Minho region (Viana do Castelo,
Ponte de Lima, Valença, and Melgaço). To reinforce the links between all IPVC
campuses and engage the IPVC community towards sustainable development, it
is strategic to put forward the IPVC Smart & Sustainable Campus, also referred
to as IPVC-S2C, which will allow not only the engagement of all IPVC commu-
nity and staff(students, professors, researchers, and employees) in all process
but will also secure the commitment and participation of the managerial officers
as a whole.
The general ICT infrastructure of the IPVC-S2C is shown in Fig. 1,inwhich
the overall digital ecosystem with all included ICT infrastructure elements and
their use is properly categorized. Moreover, the relations of the different ICT
infrastructure elements and their interactions are also identified.
The foundation of the IPVC-S2C ICT infrastructure is based on the IoT
Edge Devices, such as sensors and actuators that are being used in three main
application categories: 1) Smart Metering, which includes emissions, water dis-
tribution, and renewable energy production management, 2) Smart Mobility,
including the context information management of a network of electric bicy-
cles (BiRa Project), and 3) Smart Building, which includes Indoor Air Quality
assessment in the classroom and access/attendance control. These devices will
communicate the acquired data via WiFi, Cellular 3G/4G, or LoRaWAN, which
will reach the IoT Agent that creates a standardized interface to all IoT inter-
actions, allowing the data to be managed by the FIWARE Context Broker.
After context managing and if some predefined events have been detected,
authenticated users, such as Staff, Professors, and/or Students can be notified
to take actions, or requests can be made (HTTP, MQTT, etc.) to the IPVC-S2C
560 P. Martins et al.
General
Network
General
Network
Radio Access
Technology
(RAT)
Fiware
Generic
Enablers
(FGE)
Operations,
Administration &
Management (OAM)
IoT Edge Devices
(Sensors, Actuators)
(IoT-ED) Smart Metering
IoT Agents IoT Agents
Research &
Learning (RL)
Staff StudentsProfessors
OAM APIs (REST, etc)
Cellular 3G/4G
NB-IoT
Internet
Internet
Authentication
Server
IPVC Smart Campus
Application Server Other OAM
Servers
Smart Building
Smart Mobility
Context Broker
Identity Manager
Ethernet, Optical Fiber
Ethernet, Optical Fiber
Fig. 1. IPVC-S2C ICT infrastructure.
Application Server or other OAM Servers, so the applications change the behav-
ior (automatically), triggering smart decisions within the context of a Smart and
Sustainable Campus.
Before using FIWARE, several platforms had been developed by using a
monolithic architecture, allowing services to be deployed as a single solution.
Given the fact that these are small applications, their development was easier to
be achieved. However, being scalability a priority in the current days and since
monolithic applications are difficult to modify, a new approach was required.
Based on the related works and having in mind the application requirements
and the methodology, the proposed architecture is conceptually split into two
major building blocks:
a) IoT Devices with Edge-enabled Intelligence;
b) IPVC-S2C Digital Ecosystem
1. FIWARE Generic Enablers;
2. Short-Term Historic and Real-Time Data;
3. Application Server.
The data will be collected and transmitted by several IoT Edge devices that are
being used in several monolithic projects inside the campus. Communications
using WiFi or LoRaWAN networks are used for Smart Metering and Smart
Building applications, and LoRaWAN or Cellular 3G/4G for Smart Mobility
applications, as shown in Fig. 1.Thefollowingtwosections(Sect.4and 5)will
be used for a detailed description of the two main building blocks previously
identified.
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 561
4 IoT Devices with Edge-Enabled Intelligence
Edge devices can sense, measure, interpret, and transmit data up to the cloud
through an internet gateway. In cloud-centered architectures, the raw data is
pushed to a centralized server by the end device without any type of processing.
However, IoT Edge devices are becoming more efficient, affordable, and power-
ful, which enables low-latency real-time processing and the distribution of the
computational cost between the IoT Edge devices.
Recently, the fusion of Artificial Intelligence (AI) and edge computing is
becoming a reality. On one hand, AI intends to implement intelligent human
behavior in devices/machines by extracting knowledge and learning from data.
On the other hand, edge computing aims at coordinating a multitude of collabo-
rative edge devices and servers to process the generated data in proximity to the
source of data. Another relevant requirement that must be taken into account
when considering edge-enabled Intelligence is the low-latency, which can be guar-
anteed by one of the three relevant architectures relies on Deep Neural Networks
(DNNs) that are executed at the end device. Alternatively, Edge server–based
computation relies on data that is sent by the end devices to edge servers for
computation. Lastly, Joint computation includes the possibility of having cloud
processing.
However, bringing AI to the edge is a challenge due to the limited resources
available in common hardware used to design IoT edge devices. One approach
that has been implemented with relative success is to reduce the model’s infer-
ence time. To run an AI model in an embedded IoT device, the hardware needs
to be properly selected to fit the model design and compression [6]. Edge AI is
normally based on models with a reduced number of parameters in the Deep
Neural Network (DNN) model, which considerably reduces memory needs and
execution latency while preserving high accuracy.
Model compression, i.e. reduction of the model size, can be achieved using
quantization and pruning techniques, individually or working together. Post-
training quantization reduces computing power demand and energy consumption
at the expense of a slight loss in accuracy, allowing to run the model on tiny
devices. On the other hand, pruning eliminates non-essential connections for the
Neural Network (NN) and consequently reduces (1) the number of computations
and (2) the demand for memory space for the NN [6].
Hardware selection must be based on the analysis of 4 relevant metrics: 1)
Accuracy; 2) Energy efficiency; 3) Throughput; 4) Cost. The accuracy of the
machine learning algorithms must be quantified using large data sets to guar-
antee that the obtained results are valid. Energy efficiency is a metric directly
related to the adaptation of the model to the context change, i.e., the model
adapts its weights as the scenario changes, which involves recurrent memory
access for reading/writing weight values, resulting in increased energy consump-
tion. The throughput metric represents the number of operations required per
unit of time and the cost is directly related to the amount of memory required
to host the model, because memory is still the critical building block in com-
putational systems and although model compression can be applied, the model
562 P. Martins et al.
size can still be in the order of tens or hundreds of megabytes, heavily impacting
the overall edge device cost. Table 1presents a selection of hardware commonly
used in the design of IoT devices that can perform edge computing and specifi-
cally run Machine Learning (ML) Libraries, such as Tensorflow Lite. Tensorflow
Lite is an example of an ML library specifically implemented to be used in
microcontroller-based constrained edge devices [7].
Table 1. IoT hardware compatible with TensorFlow Lite. Adapted from [6].
Board Processor Power Connectivity Cost
Arduino Nano 33
BLE Sense [8]
ARM Cortex-M4
32-bit@64 MHz
52 µA/MHz BLE €27
SparkFun
Edge [9]
ARM Cortex-M4F
32-bit@48/96 MHz
6µA/MHz BLE 5 €15
Adafruit
EdgeBadge [10]
ATSAMD51J19A
32-bit@120 MHz
65 µA/MHz BLE/WiFi €35
ESP32-
DevKitC [11]
Xtensa dual-core
32-bit@160/240 MHz
2mA/MHz BLE/WiFi €10
Figure 2depicts the generic architecture of the IoT Edge device with its core
elements identified—sensors and actuators, microprocessor, and communication
radios—which have been in use in several application domains, cf. [18–22], within
the IPVC S2C Digital Ecosystem.
MCU
ACTUATOR
SENSOR 1
SENSOR 2
µProcessor
Power Management
Hardware Security
Sensors
and Actuatos
Wi-Fi
Radio
LoRa
Radio
HTTPs Secure
Communications
AES Encrypted
Communications
Signal Processing
and Analytics Communications
Fig. 2. IoT Edge generic architecture.
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 563
5 IPVC-S2C Digital Ecosystem
The blocks that are the base of the IPVC-S2C Digital Ecosystem are illustrated
in Fig. 3.Thisapplication-orientedarchitecturewasdesignedusingasetof
FIWARE Generic Enablers (GEs) that interact with the IoT Edge devices and
other third-party systems through a context broker. Short-Term historic and
Real-Time pre-processed data will be stored and aggregated in a time-series
database (TSDB). TSDB can handle large amounts of data while delivering
fast response due to its native optimization for storing and querying time series
data, enhancing the data compression rate and the data manipulation speed
when real-time analytics over large sets of timestamped data is required [12].
AppDB
MongoDB
Keyrock
Identity
Manager
Client APP
FrontEnd
Orion
Context
Broker
IoT
Agent
IPVC S2C Digital Ecosystem
Grafana
IoT Edge 0
IoT Edge N
. . .
App
Server
External APIs
Services
Wi-Fi
Access
Point
IoT Edge 1
LoRaWAN
Server
Short-Term Historic
& Real-Time Data
CEP
Complex Event
Processing
IoT Edge Devices
FIWARE App Server
RESTful
API
TSDB
InfluxDB
IPVC S2C App Server
User
Authentication
Device
Authentication
Legend:
Application Data
Time Series Data
Fig. 3. Overall System Architecture including the FIWARE Generic Enablers and other
third party services and modules.
In the proposed architecture, four FIWARE GEs were used. The Orion Con-
text Broker, allows the register, update, and query operations of the context data
and working with publish/subscribe communication patterns, via notifications to
the responsible organization. The Keyrock Identity Management module enables
identity management and authentication/authorization security to the services
and applications. The IoT Agent allows data transfer between the sensors and
the Context Broker. The Complex Event Manager (CEP) analyses event data
in real-time and enables instant responses to change conditions, such as notifi-
cations, emails, tweets, and messages [13].
564 P. Martins et al.
The TSDB directly connects to the Context Broker and the IoT agents and
can serve data directly to the client App through Grafana as-a-service, enabling
the on-the-fly generation of rich and interactive dashboards. Since external ser-
vices can be deployed independently, and since other already existent database
legacy systems are in use on the campus they will also be integrated through
external APIs and Webservices, such as RESTful APIs. In Subsects. 5.1 to 5.3,
each building block of the IPVC-S2C Digital Ecosystem will be introduced and
explained in detail.
5.1 FIWARE App Server
The main core (and only needed) of any FIWARE-based application is the
Orion Context Broker. According to [14], Orion Context Broker is the Pub-
lish/Subscribe implementation, decoupling consumers data, functioning based
on an Open API Next Generation Service Interface (NGSI), on two versions,
NGSI-9 and NGSI-10, which defines the data model, context data interface, and
availability [15], that allow the registration, updates, queries and notifications of
context data.
To analyze and pro cess data in real - t i m e , a C o m p l e x E vent Processing (CEP),
also referred to as event stream analysis, or real-time event correlation, will
be installed, allowing immediate response to changing conditions, like sending
emails, SMS messages, HTTP requests, tweets, etc. The CEP API allows the
management of rules, exposing CRUD operations, being triggered by feeding
them with NGSI10 notifications [13].
For security m e a s u r e s, the Keyro ck Identity M a n a g e r w i l l b e u s e d , w h i ch will
be responsible for authentication and providing access to information, offering a
Graphic User Interface based (GUI-based) or API-based interaction to admin-
istrative users, roles, and permissions enabling the addition of OAuth2-based of
users and devices, user profile management, privacy-preserving of personal data,
Single Sign-On and Identity Federation across multiple domains, enabling the
register of OAuth 2.0 consumers as Service Providers [16]. It is used to create a
secure FIWARE application and contains data of Users - any human actor inter-
acting with a FIWARE application and Organizations - An association allowing
certain users to administer all rights. The Identity Manager ensures that only
the right individuals get access to resources, such as usernames, passwords, and
roles, and the access control is the selective restriction of access to resources, with
Authorization and Authentication. The Identity Manager is key in the architec-
ture since it reduces the work on account creation and management, using the
user profile storage as a Software as a Service, supporting the usage of policies
and procedures. The administrators can easily configure access to services and
the handling of error notifications. Since it re-uses attribute data, it allows easy
and convenient management of profile information [13].
An IoT Agent is the GE that makes it possible for a group of devices to
send their data to be managed from a Context Broker. It translates an IoT
specific protocol into an NGSI v2, overcoming common problems in the IoT
domain, such as mapping the data received in a meaningful manner, abstracting
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 565
communications so users can remain unaware of the device-specific protocols,
bringing a standard interface to all IoT interactions at the context information
level. The IoT Agent supports a single message format and can be configured to
use transports such as HTTP, MQTT, and AMQP [13].
5.2 Short-Term Historical and Real-Time Data
The Short-Term Historic and Real-Time Data block of the Smart Campus archi-
tecture is composed of a single time-series database, InfluxDB, an open-source
database designed to handle high write and query loads, which will be used to
store the data sent by the IoT Edge Devices. Since the Smart Campus will use
multiple sensors, InfluxDB was chosen to store multiple data collected by the
sensors, because it is fast and scalable, supporting millions of writes per second,
having the ability to handle specific functions to accelerate data processing [17].
Historical context data can be persisted to the InfluxDB, resulting in a series
of data points, which are meaningless on their own but combined can be trans-
formed into meaningful statistics, which can be displayed, with an easy user
interface and enhanced visual analytics to the User as Dashboards or metrics
and KPIs for distinct periods: Real-Time (last hour), Short-Term (last 7 days),
using Grafana-as-a-service in the Front End.
5.3 IPVC-S2C App Server
The IPVC-S2C App Server is conceptually a block composed of two main cores
that can be described as the external services and the application server, which
can be considered as a Back End to the IPVC Management Front End and the
AppDB. The App Server will be used as an external API, allowing the connection
to the External API Services for the existing Core Applications, such as ICT’s,
Nature and Environment, Emissions, Water and Energy, Mobility, Health, and
Research and Innovation. The AppDB is a MongoDB open-source database,
which is dynamic and object-oriented, having high performance, availability, and
automatic scaling, and will be used to the Back End data that already exists
from the monolithic approach.
6PreliminaryResults
The proof of concept used to test the IPVC S2C digital ecosystem was set up
using an IoT Edge device designed for Indoor Air Quality monitoring in the
schools, cf. [22]. The device was designed to monitor Indoor Air Quality and
collect parameters such as particle matter (PM1.0, PM2.5, and PM10), total
volatile organic compounds (TVOC), CO2concentration, and air temperature
and relative humidity.
Figure 4depicts the IAQ4Classroom client application, a cartography based
web-application platform centered around a map, using Leaflet and Geoserver,
two geographic information system (GIS) tools to enhance visual data analytics
566 P. Martins et al.
Fig. 4. Map-centered front end with related IoT Edge device dashboard.
that allows high-level building management using native and geo-referenced hier-
archies between entities, i.e., school > floor > classroom > sensor, hierarchies
defined and obtained using the FIWARE reference data context model (NGSI
v2), to perform spatial queries, by entity or entity aggregate. These entities are
implemented as GeoJSON vectors and are assigned a specific color that changes
in real-time, related to a risk indicator or a legal limit. For an easier user data
visualization, Grafana will be used to display Dashboards, KPIs, and Metrics
according to the collected data in three time periods: Very Short-Term for peri-
ods of 24 h, Short-Term for periods of 1week to three months, and Long-Term
for periods of 3 months plus.
With the implementation of FIWARE Orion Context Broker, operations
related to the entity state, such as create, query and update are basic for syn-
chronous context producer and consumer applications. However, Orion Context
Broker has a strong advantage of letting the user know the information as soon
as it arrives since it allows the ability to subscribe to context information so
the application gets an asynchronous notification when “something” happens,
enabling a faster and better deployment since it removes polling. In addition to
the Broker, the CEP analyses data in real-time to generate an instant response,
manual or automatic, to changing conditions, such as notifications to the end-
user. The IoT Agent is automatically connected to the IAQ4Classroom sensor
and the corresponding data with specific content in the Orion Context Broker,
storing the devices’ configuration in the AppDB. In our use case scenario, each
time the measurement is collected, the data is updated by the Orion Context
Broker.
7 Conclusions and Future Work
The current research shows the design methodology, which serves as the ground
for the specification of a FIWARE-based Smart & Sustainable Campus with
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 567
Edge-enabled Intelligence. At the time of writing, the project is being tested
using a LoRa-based IoT Edge device to measure the Indoor Air Quality, within
the Smart Building category, as presented in Fig. 1.FIWAREalsoallowsthe
inclusion of several modules that are not displayed in the proposed architecture,
cf. Fig. 3, that may be added if necessary, such as the CKAN extension, an open
data publication module that allows the publication of data-sets and assignment
of terms, policies, pricing and pay-per-use schemes to data-sets [13]. The use
of FIWARE technologies enables simplifies the design of new applications and
streamlines its integration with other legacy systems that are already in use
on the campus. Moreover, the use of a context broker centralizes and provides
context to data that comes from different IoT Edge devices. In this approach,
external entities can collect, process, and display information without needing
to directly interact with data sources.
Future wo r k w i l l i n c l u d e t he integration of other existing m o n o l i t h i c a p p l i -
cations that are still in use on the campus, mainly concerning the project
Refill_H20, which proposes to reduce the sale of plastic water bottles in Poly-
technic Institute of Viana do Castelo (IPVC), a higher education institution,
to promote the circular economy, by reducing plastic materials use and waste.
By promoting the reduction in disposable waste production, project Refill_H20
will help to reduce energy consumption and greenhouse gases emission. The
ICT infrastructure for the IPVC-S2C is still under development, lacking bench-
mark tests to the IoT Edge devices for the Smart Metering and the Smart
Mobility application categories, however, the application to Project Refill_H20
will help to boost its implementation. Context data notifications will be also
implemented to advise the responsible user or organization or to automatically
trigger actions in response to certain physical or environmental conditions. New
FIWARE Generic Enablers will also be implemented according to the needs of
the community. Usability tests should be developed and deployed for the Front
End application to evaluate the application flow and improve its user interface.
The goal of these tests is the simplification of the user interaction and the per-
ceptibility of the data acquired by the sensors.
Acknowledgments. The authors wish to thank especially the Program Environment,
Climate Change and Low Carbon Economy, created following the establishment of
a Memorandum of Understanding between Portugal and Iceland, Liechtenstein and
Norway, under the EEA and Norway Grants 2014–2021, for the program areas of
Environment and Ecosystems (PA11), and Climate Change Mitigation and Adaptation
(PA13), for financing the project 10_SGS#1_REFILL_H20.
References
1. SmartCAMPUS UWB, SmartCAMPUS. https://www.smartcampus.cz/.Accessed
04 May 2020
2. WiseTown, Smart City Project. https://wise.town/. Accessed 04 May 2020
3. Smart Metropolis. http://smartmetropolis.imd.ufrn.br/. Accessed 06 May 2020
568 P. Martins et al.
4. Leonardo Santiago, Bettair. https://www.fiware.org/2018/08/26/fiware-solution-
bettair-on-a-mission-to-map-urban-air-quality-in-high-resolution/.Accessed14
May 2020
5. Fortes, S., Santoyo-Ramón, J.A., Palacios, D., Baena, E., Mora-García, R.,
Medina, M., Mora, P., Barco, R.: The campus as a smart city: university of Málaga
environmental, learning, and research approaches. Sensors 19,1349(2019)
6. Merenda, M., Porcaro, C., Iero, D.: Edge machine learning for AI-enabled IoT
devices: a review. Sensors 20,2533(2020).https://doi.org/10.3390/s20092533
7. Tensor Flow Light Guide. https://www.tensorflow.org/lite/guide. Accessed 14 July
2020
8. Arduino Nano 33 BLE Sense. https://store.arduino.cc/arduino-nano-33-ble-sense.
Accessed 14 July 2020
9. SparkFun Edge Development Board - Apollo3. Blue. https://www.sparkfun.com/
products/15170. Accessed 14 July 2020
10. Adafruit EdgeBadge - Tensorflow Lite for microcontrollers. https://www.adafruit.
com/product/4400. Accessed 14 July 2020
11. Espressif Systems ESP32 Overview. https://www.espressif.com/en/products/
devkits/esp32-devkitc/overview. Accessed 14 July 2020
12. Jensen, S.K., Pedersen, T.B., Thomsen, C.: Time series management systems: a
survey. IEEE Trans. Knowl. Data Eng. 29(11), 2581–2600 (2017). https://doi.org/
10.1109/TKDE.2017.2740932
13. FIWARE Foundation, Developers Catalogue. https://www.fiware.org/developers/
catalogue. Accessed 15 May 2020
14. Araujo, V., Mitra, K., Saguna, S., Åhlund, C.: Performance evaluation of FIWARE:
a cloud-based IoT platform for smart cities. J. Parallel Distrib. Comput. (2019).
https://doi.org/10.1016/j.jpdc.2018.12.010
15. Ivan, T., Lalovic, K., Trikos, M.: FIWARE: a web of things development platform.
Military Technical Courier (2018). https://doi.org/10.5937/vojtehg66-17063
16. Alonso, A., Pozo, A., Choque, J., Bueno, G., Salvachúa, J., Diez, L.: An identity
framework for providing access to FIWARE OAuth 2.0-based services according
to the eIDAS European regulation. IEEE Access 7,88435–88449(2019).https://
doi.org/10.1109/ACCESS.2019.2926556
17. InfluxData Inc, InfluxDB. https://www.influxdata.com/products/influxdb-
overview/. Accessed 15 June 2020
18. Lopes, S.I., Moreira, P.M., Cruz, A.M., Martins, P., Pereira, F., Curado, A.:
RnMonitor: a WebGIS-based platform for expedite in situ deployment of IoT edge
devices and effective radon risk management. In: IEEE International Smart Cities
Conference (ISC2), Casablanca, Morocco 2019, pp. 451–457 (2019). https://doi.
org/10.1109/ISC246665.2019.9071789
19. Lopes, S.I., Pereira, F., Vieira, J.M.N., Carvalho, N.B., Curado, A.: Design of com-
pact LoRa devices for smart building applications. In: Afonso, J., Monteiro, V.,
Pinto, J. (eds.) Green Energy and Networking. GreeNets 2018. Lecture Notes of the
Institute for Computer Sciences, Social Informatics and Telecommunications Engi-
neering, vol. 269, pp. 142–153. Springer, Cham (2019). ISSN 1867-8211. https://
doi.org/10.1007/978-3-030-12950-7_12
20. Pereira, F., Lopes, S.I., Carvalho, N.B.: Design of a cost-effective multimodal IoT
edge device for building occupancy estimation. In: IEEE International Smart Cities
Conference (ISC2), Casablanca, Morocco 2019, pp. 122–128 (2019). https://doi.
org/10.1109/ISC246665.2019.9071717
Designing a FIWARE-Based Smart Campus with Edge-Enabled Intelligence 569
21. Pereira, F., Lopes, S.I., Carvalho, N.B., Curado, A.: RnProbe: a LoRa-enabled IoT
edge device for integrated radon risk management. IEEE Access 8,203488–203502
(2020). https://doi.org/10.1109/ACCESS.2020.3036980
22. Abreu, A., Lopes, S.I., Manso, V., Curado, A.: Low-cost LoRa-based IoT edge
device for indoor air quality management in schools. In: EAI Edge-IoT, EAI Inter-
national Conference on Intelligent Edge Processing in the IoT Era, 2–4 December
2020. N/A, Cyberspace (2020)
