Content uploaded by Jeroen Famaey
Author content
All content in this area was uploaded by Jeroen Famaey on Sep 11, 2017
Content may be subject to copyright.
Real-time Data Dissemination and Analytics
Platform for Challenging IoT Environments
Glenn Daneels∗, Esteban Municio∗, Kathleen Spaey∗, Gilles Vandewiele†,
Alexander Dejonghe†, Femke Ongenae†, Steven Latr´
e∗, and Jeroen Famaey∗
∗University of Antwerp – imec, Belgium, IDLab, Department of Mathematics and Computer Science
†Ghent University – imec, Belgium, IDLab, Department of Information Technology
Abstract—The advent of Internet-of-Things (IoT) applications,
such as environmental monitoring, smart cities, and home au-
tomation, has taken the IoT concept from hype to reality at a
massive scale. However, more mission-critical application areas
such as energy, security and health care do not only demand
low-power connectivity, but also highly reliable and guaranteed
performance. While fulfilling these requirements under controlled
conditions such as urban and indoor environments is relatively
trivial, tackling the same obstacles in a more challenging and dy-
namic setting is significantly more complicated. In environments
where infrastructure is sparse, such as rural or remote areas,
specialized infrastructure-less ad-hoc solutions are needed, which
provide long-range multi-hop connectivity to remote sensors and
actuators. In this paper we propose a new general-purpose IoT
platform based on a combination of Low-power Wireless Personal
Area Network (LoWPAN) and multi-hop Wireless Sensor Net-
work (WSN) technology. It supports reliable and guaranteed real-
time data dissemination and analysis, as well as actuator control,
in dynamic and challenging infrastructure-less environments. In
this paper, we present the IoT platform architecture and an initial
hard- and software prototype. Moreover, a use case based on real-
time monitoring and training adaptation for cyclists is presented.
Based on this case study, evaluation results are presented that
show the ability of the proposed platform to operate under
challenging and dynamic conditions.
I. INTRODUCTION
The Internet-of-Things (IoT) paradigm is rapidly maturing,
with many massive deployments in areas such as smart cities,
home automation, and agricultural monitoring. These massive
IoT applications require scalable low-power connectivity over
a long range. Recently, there has been a rising demand for
IoT in more critical sectors, such as health care, security,
energy and industrial automation. They are referred to as
mission-critical IoT applications, and additionally require spe-
cific Quality-of-Service (QoS) guarantees in terms of latency,
throughput or reliability.
Current IoT deployments generally use Low-Power Wide
Area Network (LPWAN) or cellular technologies (e.g., LoRa,
Sigfox, NB-IoT, GPRS) to provide (low-power) connectivity
to sensors and actuators. However, such solutions require an
(often public and operator-managed) infrastructure of base
stations or access points with sufficient coverage. This makes
them highly suitable to provide low-cost best-effort connec-
tivity in easily accessible urban environments. However, in
rural or remote areas, this infrastructure is not always present.
Moreover, these technologies generally do not support QoS
differentiation.
Low-power Wireless Personal Area Network (LoWPAN)
and Wireless Sensor Network (WSN) technologies, such as
IEEE 802.15.4, are far more suited for such challenging en-
vironments, as they can provide infrastructure-less long-range
connectivity over multiple hops. Several recent amendments
have been made to the 802.15.4 standard to improve its suit-
ability for IoT. IEEE 802.15.4e proposes an enhanced MAC
layer based on Time-Slotted Channel Hopping (TSCH) that
greatly increases reliability and efficiency in difficult and dense
settings, such as industrial environments. IEEE 802.15.4g
proposes a PHY amendment for sub-1GHz operations, which
greatly increases the range of a single hop from dozens to
hundreds of meters, and improves energy efficiency. The com-
bination of IEEE 802.15.4e and g therefore has the potential
to enable long-range low-power multi-hop connectivity with
reliability and QoS guarantees in challenging and remote areas.
In this paper, we propose an end-to-end IoT platform for
real-time sensor data dissemination and analytics, as well as
actuator control in challenging and remote areas. It builds
upon IEEE 802.15.4e and g to provide multi-hop long-range
connectivity between sensors and data sinks. The TSCH
MAC scheduler is optimized to significantly reduce latency
to support real-time data analytics. Moreover, the platform
provides multi-modal communication with Bluetooth Low
Energy (BLE) and ANT+ to support the connection of a
variety of wireless sensors and actuators. Finally, integrated
distributed data analytics features support real-time network-
aware data processing either inside the network or in the cloud
back-end when an Internet connection is available. In order to
exemplify the benefits of the platform, we present a real-life
case study. It offers a real-time data-driven cycling experience,
where amateur or professional cyclists wear a set of sensors.
This allows intelligent agents or coaches to monitor them and
provide real-time feedback and continuous training adaptation.
Evaluation results based on a prototype implementation of
the platform show its advantages for challenging IoT envi-
ronments.
The remainder of this paper is organized as follows.
Section II presents related work in the area of IoT data
dissemination and analytics platforms. Section III introduces
the proposed IoT platform for real-time data dissemination
and analytics in challenging areas. Subsequently, Section IV
introduces a representative cycling monitoring and training
adaptation use case on which the prototype implementation
was applied. The preliminary prototype and initial evaluation
results are presented in Section V. And finally, Section VI
presents future research challenges and our conclusions.
II. RE LATE D WORK
This section presents related work in two related IoT areas:
(i) providing low-power connectivity for real-time data dissem-
ination in challenging environments, and (ii) performing real-
time data analytics on top of such a dissemination platform.
A. IoT Connectivity in Challenging Environments
Most of the research on IoT platforms is focused on Smart
Cities, where non-resource constrained networks monitor or
control city assets [1], [2], [3]. Usually these approaches are
based on wireless technologies that are highly dependent on
a dense infrastructure of interconnected base stations and are
therefore not suitable for rural, remote and challenging areas.
On one hand, as a potential solution, using long-range
LPWAN or cellular technologies for providing rural con-
nectivity has been proposed by several researchers [4], [5],
[6]. Examples include the use of traditional mobile networks
(e.g., GPRS, LTE), NarrowBand-IoT (NB-IoT) [7], LoRa,
and SIGFOX. Due to the high range of tens of kilometers
of such technologies, these solutions can rely on a very
sparse infrastructure of base stations attached to a high-speed
back-haul network. While NB-IoT offers the best QoS, LoRa
and SIGFOX are technologies usually used only for low-
throughput and delay tolerant applications. These technologies
have the disadvantage that deploying and maintaining base
stations is expensive and extending the network infrastructure
to increase coverage is only economically viable if enough
potential customers are present in the area. Rural, remote
and challenging environments with low population density are
therefore not profitable and will often not have (sufficient)
coverage.
On the other hand, infrastructure-less approaches have been
presented [8], [9], [10]. These approaches generally extend
coverage through inexpensive multi-hop or meshing protocols.
The technologies used can vary from broadband solutions such
as Wi-Fi to low-power multi-hop WSN technologies such as
DASH7 [11], Zigbee [12], and more recently IPv6 over the
TSCH mode of IEEE 802.15.4e (6TiSCH) [13].
In this paper, we propose a hybrid solution that combines the
advantages of infrastructure-based and infrastructure-less low-
power connectivity for IoT in a single multi-modal platform.
It employs an ad-hoc sub-1GHz wireless multi-hop network
for providing end-to-end connectivity between sensor devices
and the sinks. For this purpose, 6TiSCH together with IEEE
802.15.4g is used, enabling low-power, long-range, and re-
liable data dissemination. Moreover, a variety of LoWPAN
technologies (e.g., BLE [14] and ANT+[15]) are provided by
each device to connect a plethora of heterogeneous sensors and
actuators. Using BLE, the sink device can connect to a smart-
phone or tablet, which in turn provides Internet connectivity
and cloud synchronization using Wi-Fi or LTE whenever they
are available. The proposed multi-modal platform provides
a highly versatile IoT solution for challenging environments
compared to state of the art solutions, with infrastructure-
less data dissemination, sporadic disruption-tolerant Internet
uplink, and support for heterogeneous wireless sensors and
actuators.
B. Real-time Data Analytics for IoT
Semantic web technologies are often adopted to consolidate
the heterogeneous and voluminous IoT data streams and
expose them to the data analytics components in a uniform
manner [16], [17]. They impose a common data representation,
make the properties of the device and the context in which the
data was gathered explicit and enable the straight-forward in-
tegration with background knowledge, e.g., the medical profile
of patients or the layout of a smart city. Semantic reasoning
enables to derive actionable insights out of this interlinked
data, e.g., the light in a room is too bright for a patient with
concussion. Several semantic IoT middleware frameworks
have emerged to enable (real-time) semantic data annotation
and support the development of data analytics components
based on this enriched data [18], including Sense2Web [19],
XGSN [20], [21], and SOFIA2 [22]. However, most of them
assume centralized processing of the data in the cloud. They do
not take into account the support for real-time local decision
making for critical applications in which Internet connectivity
and resources might be sparse. In contrast, we propose a
solution that provides real-time distributed reasoning and data
analytics inside the network, in combination with sporadic
cloud-based offloading.
III. IOT PL ATFO RM F OR CHALLENGING ENVIRONMENTS
In this section, an end-to-end IoT platform for real-time
sensor data dissemination and analytics, as well as actuator
control in challenging and remote areas, is presented. Ex-
amples of such harsh environments include mountain trails
or sparsely populated rural areas which lack static cellular
infrastructure or other long-range IoT technologies, such as
GPRS, LTE, NB-IoT, LoRA, or SIGFOX. We target dynamic
network scenarios in which IoT devices are attached to mobile
assets (e.g., vehicles, bicycles, or animals). In many such use
cases the platform has to support real-time data dissemination
among up to hundreds of power-constrained, wireless nodes
in a limited geographical area, which all contend for the
same wireless spectrum. A schematic representation of the
proposed platform to handle these challenges is presented in
Figure 1. The remainder of this section describes the different
components of this platform.
A. Device Functionality
The platform will contain two types of devices that each
have specific tasks: nodes and sinks. Platform nodes feature
sensors and actuators, disseminate collected data and can
Sink Node BLE
IEEE 802.15.4g ANT+
Data Analysis Sporadic connection to the cloud
Actuator
Actuator
Sensor
Sensor Sensor Sensor
Sensor
Sensor
Fig. 1: An example of the IoT platform with sensors and
actuators attached to mobile devices connected to one or
more sinks over multiple hops, distributed data analytics
components that provide real-time analysis, actuator feedback
and visualization (through a connected smartphone or tablet).
Sporadic cloud uplink is also provided through a connected
smartphone or tablet device.
perform elementary data processing. Platform sinks receive
monitored data from the network and provide almost instant
feedback to the individual nodes after applying advanced data
analytics.
1) Platform node: The functionality of an individual node
in the network is threefold, as seen in Figure 1. First, it
connects to both sensors (e.g., heart rate sensor or GPS tracker)
and actuators that, respectively, send periodically captured data
to the node and respond to received feedback from a data
analytics unit in the network. Second, it is also responsible to
relay monitored data towards the platform sinks and receive
data coming from those sinks, using a multi-hop path as further
explained in Section III-B. Third, it features elementary data
processing capabilities that can provide real-time feedback to
users or to actuators.
An important requirement is low-power consumption as
these platform nodes are typically designed to be very small
to avoid being a burden to the carrier. To be energy-efficient,
only (ultra) low-power wireless protocols designed for for-
warding sensor data, such as ANT+ or BLE, are considered.
Also, wired connectivity is supported. However, this is often
physically inconvenient or simply not feasible.
2) Platform sink: The platform sink receives the collected
data from the network nodes and can responds with real-
time feedback that will drive the actuators connected to the
nodes. Therefore, the sink does not connect to any sensors
itself, but is rather a collector of all monitored data or a
subset of the data (as there can be multiple sinks in the
network), as shown in Figure 1. The sink provides near-instant
feedback to the nodes based on local data processing, data
processing on a peripheral device, or in the cloud back-end
when uplink connectivity is available. The connection between
the sink and a peripheral device, such as a smartphone or
tablet, can be any wireless protocol as the sink typically will
have sufficient power resources compared to the other platform
nodes (e.g., using Wi-Fi or BLE). The peripheral device can
perform advanced data analytics on the received data and/or
can use a visualization application to show the data to users.
If available, cloud resources could also be used to analyze the
data, but a steady Wi-Fi or 3G/4G connection is not guaranteed
in challenging and remote environments and will rather be
sporadic.
B. Data Dissemination
The platform should support the end-to-end dissemination
of real-time data from sensors to the sink and in the reverse
direction over a range of up to several kilometers, without
relying on fixed network infrastructure such as 4G or LPWAN.
Due to transmit power constraints and possible lack of line-of-
sight between source and sink, end-to-end connectivity will re-
quire multi-hop communication, as shown in Figure 1. There-
fore, a multi-hop network is set-up using the IEEE 802.15.4g
physical layer in combination with the IEEE 802.15.4e MAC
layer. This turned out to be the most promising solution after
comparing different infrastructure-less and wireless candidate
technologies (e.g., Wi-Fi, BLE, DASH7, and IEEE802.11ah).
In terms of satisfying throughput and distance constraints,
low-power operations and multi-hop support, it surpassed all
other existing technologies. IEEE 802.11ah also seemed very
promising but its current lack of hardware makes its use in
the platform impossible. IEEE 802.15.4g is a recent PHY
amendment to the IEEE 802.15.4 standard which supports an
infrastructure-less mode and is specifically tailored for low-
power long-range communication by using the 868 MHz band.
Another advantage of the IEEE 802.15.4g physical layer is that
it can be combined with the so-called 6TiSCH architecture
that combines industrial performance in terms of reliability
and power consumption with a full IPv6-enabled IoT upper
stack [13]. Two important aspects of the 6TiSCH architecture,
i.e., schedule management and routing, are discussed in more
detail.
1) Schedule management: 6TiSCH uses the TSCH mode
of the IEEE 802.15.4e MAC layer which combines channel
hopping to avoid external interference and multi-path fading
with a TDMA-based schedule that allows network nodes to
be extremely energy-efficient and reliable. The tight time-
synchronized TX/RX schedule avoids battery power waste
by allowing nodes to turn their radio off and only turn it
on when they are expected to send or receive data during a
specified time slot. A scheduling algorithm is responsible for
the management of this schedule by intelligently assigning
Fig. 2: Cascaded reasoning pipeline to enable real-time semantic reasoning within environments with sparse Internet connectivity
and limited local resources.
available time-frequency resources to nodes that have data
ready for transmission.
The 6TiSCH Working Group (WG) is standardizing a
scheduling function called Scheduling Function Zero (SF0)
which dynamically adapts the number of reserved time slots
between neighbor nodes, based on the currently allocated
bandwidth and the requirements of a node’s neighbors [23].
SF0 however performs no intelligent allocation of specific
time slots: it picks its time slots randomly which can lead
to suboptimal latency results. A more intelligent approach of
scheduling time slots is used by the Low Latency Scheduling
Function (LLSF) which daisy-chains receiving and transmit-
ting time slots in a multi-hop path in order to decrease the
latency [24]. We implemented an enhanced version of LLSF,
called Enhanced Low Latency Scheduling Function (eLLSF)
that explicitly differentiates between reserving time slots to
a parent or to a child, a feature that was not discussed in
the original work on LLSF. When eLLSF reserves multiple
TX time slots to a parent, it will distribute these time slots
evenly among the reception slots (with the so-called largest
gaps between the reception and transmission slot) of all the
children. If not all time slots can be distributed evenly, the
remainder of the time slots is allocated randomly to the
children. When reserving a time slot to a child, the original
three-step reservation process of LLSF is used.
2) Routing: To route the data throughout the network to the
appropriate sinks, the IPv6 Routing Protocol for Low-Power
and Lossy Networks (RPL) is used [25]. RPL was specifically
designed to face the typical challenges of WSNs i.e., mini-
mizing energy consumption and latency while still satisfying
multiple network constraints. In RPL, each network topology
is organized as a tree, called a Destination Oriented Directed
Acyclic Graph (DODAG). Multiple DODAG instances can
exist in a single network, relaying data to different sinks
based on different parameters. This allows the platform to have
multiple sinks and route (different subsets of) the collected
data to multiple sinks.
C. Real-time Data Analytics
The proposed platform combines stream reasoning with
the cascading reasoning paradigm [26] to support real-time
data analytics for critical applications that deal with limited
local resources and sparse Internet connectivity. Stream rea-
soning [27] focuses on the scalable and efficient adoption
of Semantic Web technologies for streaming data [28]. Re-
cently, the first prototypes of RDF Stream Processing (RSP)
engines [29] have emerged (e.g., C-SPARQL, CQELS, EP-
SPARQL and SPARQLStream). They define a window on top
of the stream and allow the registration of semantic queries
which are continuously evaluated as data flows through the
window. As such, these prototyped RSP engines can filter and
query a continuous flow of data and can provide real-time
answers to the registered queries.
As shown in Figure 2, the real-time data analytics consist
of a processing hierarchy of reasoners that exploit the trade-
off between the complexity of the reasoning and the velocity
of the data stream. At the edge of the network (i.e., on the
sink or a connected smartphone/tablet) a first filtering and
aggregation is performed on the high frequency data stream
by combining an RSP engine with one or more (traditional)
semantic reasoners. The local reasoning can be performed
more efficiently as it is done using a partial knowledge base
(i.e., only the background knowledge that is applicable to that
particular node and the devices connected to it). This enables
local decision making to rapidly act on events occurring in
the environment. It also reduces the volume and rate of the
data, as only the aggregated conclusions are forwarded to the
cloud, whenever Internet connectivity allows it. In the cloud,
more complex reasoning can then be performed that combines
conclusions derived by different edge nodes. As the data is
processed close to its source, this results in reduced network
congestions, less latency and improved scalability.
IV. USE CA SE DESCRIPTION
The IoT platform presented in Section III was evaluated
using a realistic use case in collaboration with industrial
partners, in the context of the Flemish ICON CONtinuous
Athlete MOnitoring (CONAMO) project. It encompasses real-
time collection and analysis of athlete data during collective
cycling. The goal of the use case is to improve the cycling
experience of a group of cyclists by providing them with
personalized insights in their performance.
Currently, data of cyclists, such as hearth rate, speed, power,
cadence, is captured by a multitude of devices and apps like the
ones offered by Garmin or Strava. These allow the publication
of the results of a ride after the event, but do not allow real-
time sharing of the data in environments where no adequate
cellular connectivity is available, which is often the case at
popular cycling trails (e.g., in the mountains). Several target
groups will be able to profit from a solution which offers a
real-time data-driven cycling experience. There are on one
hand the professional cycling teams, where coaches will be
able to see live performance indicators of their team members,
which they can then use to adapt training sessions or to take
strategic decisions during competition. But also during ama-
teur cycling events, receiving performance feedback or updates
about a friend’s position might enable friendly competitions
and transform cycling into a social experience.
Because of the nature of the use case, it requires many of
the features the presented IoT platform aims to realize. Each
individual bike is equipped with a node device that captures
the data of the sensors attached to the cyclist and the bike. This
data is then sent to the sink nodes, which will typically be the
team cars or motorcycles from press and jury. It is clear that
this network of platform and sink nodes is highly dynamic.
The multi-hop approach of the platform to forward the data
to the sink nodes is therefore needed because the distance
between a cyclist and a sink node might be larger than the
distance that can be bridged by a single-hop transmission.
Figure 3 shows both the distances between cyclists and sink
points for an actual Tour De France stage (stage 20 Alpe
d’Huez on July 25, 2015), for the first 50 cyclists and the
sink points (i.e., team cars and press motorcycles) in between
them. All distances are relative to the first cyclist. As the figure
illustrates, there are sometimes large gaps of up to 660 m
between the sink points (e.g., between 1.4 km and 2.1 km),
and in between these gaps there are cyclists for which the
distance to the nearest sink is several hundred meters. In a
training session of a professional team there will typically be
less sink points than in an actual race, and during amateur
cycling events the distances between cyclists are typically
larger because of bigger variations in the performance of
individual cyclists.
Because of specifics of cycling trails (presence of hills,
mountains, trees, etc.) and low transmit power settings due to
0 1 2 3 4 5
Cyclists
Sinks
Position relative to first cyclist (km)
Fig. 3: Position of cyclists and sink points (cars and motorcy-
cles) in an actual Tour de France stage.
Fig. 4: Screenshot from the CONAMO visualization app,
showing hearth rate and power output data collected from three
bikes.
battery constraints, the transmission range between two nodes
is expected to be at most a few hundred meters when using
IEEE 802.15.4g. So even with a multi-hop approach, cyclist
dynamics may cause certain nodes to temporarily become
completely disconnected from the network, motivating the
need for a disruption-tolerant routing protocol.
Once the data of a cyclist has reached the sink node, it
can be visualized, for example on the tablet of the coach
in the team car. Figure 4 shows a screenshot of a prototype
visualization app built in the context of the CONAMO project.
In order to provide real-time feedback about the on-going
training or ride to the cyclist or coach in a personalized and
context-aware fashion, the real-time data analytics part of the
platform needs to intelligently, automatically and dynamically
interpret the available data. The data set available to base the
feedback on can be very heterogeneous. There is the stream
of sensor data collected by the heterogeneous sensors on the
bikes and cyclists, but also contextual information such as the
route or weather and historical data known from each cyclist
can be taken into account. As the data analysis solution will
need to provide feedback to several cyclists at the same time,
it needs to be highly scalable.
V. PROTOT YP E & RE SU LTS
This section describes the prototype of the presented plat-
form, as developed in the context of CONAMO. The different
hardware components and their interaction are discussed. Also,
preliminary results involving the usage of IEEE 802.15.4g and
the eLLSF scheduling function are shown.
A. CONAMO Prototype
The prototype device, as shown in Figure 5, consists of
three major hardware components: a component responsible
Fig. 5: The prototype device that can serve both as node
and sink, containing an OpenMote-CC2538 and OpenUSB
combination for the multi-hop connection, a Nordic nRF52
development kit board to connect to peripheral devices and a
power bank that serves as the energy source of the device.
Sensor 1 nRF52 OpenMote
Power bank
USB
Platform node
UART
Sensor n
ANT+
BLE
Platform node
Platform sink
Node 1 OpenMote nRF52
Power bank
USB
Platform sink
UART
Node n
BLE
802.15.4g
802.15.4g
Tablet
Node n
802.15.4g
Node 1
802.15.4g
802.15.4g
Fig. 6: The prototype components and their interaction in both
a platform node and a platform sink.
for the multi-hop IEEE 802.15.4g connection, a board that
connects to other peripheral devices such as sensors and tablets
and a commercial off-the-shelf power bank that serves as an
energy source for the other two components. The different
components and their interaction are shown in Figure 6.
1) Multi-hop connectivity: To set up the IEEE 802.15.4g
multi-hop network, the prototype uses OpenMote hardware,
which was specifically designed to support full IoT stack
implementations [30]. More specifically, the prototype uses
the OpenMote-CC2538 board that carries the CC2538 system-
on-a-chip (SoC) combined with the OpenUSB board, shown
in Figure 5 [31]. While the OpenMote-CC2538 contains a
32 MHz micro-controller and a IEEE 802.15.4-compliant
2.4 GHz radio, the OpenUSB is equipped with the CC1200
radio chip that operates on the 868 MHz band. As such, our
prototype supports dual band operations in both frequency
bands. Currently, the OpenMote component in the prototype
is only used for upstream data traffic towards the sink. Future
improvements will extend the prototype functionality to also
support downward traffic (e.g., for real-time actuator control
based on data analytics conclusions).
2) Peripheral connectivity: For the BLE and ANT+ con-
nection with sensors, actuators, smartphones and tablets, we
TABLE I: Simulation parameters.
Parameter 868 MHz values 2.4 GHz values
Traffic generated (per node) 1 packet/s 1 packet/s
Packet size 51 bytes 127 bytes
Modulation 2FSK OQPSK
# Channels 2 16
# Nodes 32 32
Experiment iterations 20 20
use the Nordic nRF52 development board [32]. It is equipped
with the multi-protocol nRF52832 SoC that is targeted specifi-
cally at ultra-low power applications. This SoC supports BLE,
ANT+, IEEE 802.15.4 and proprietary 2.4 GHz protocols.
Platform nodes use BLE and ANT+ to connect to sensors,
as seen in Figure 6, as both protocols support low-power
operation and are widely supported by various commercial
sensing hardware. Meanwhile, a sink uses only BLE to connect
to tablets or smartphones, which can display the collected data
using a visualization application.
B. IEEE 802.15.4g Network Evaluation
In order to assess the ability and limitations of the 802.15.4g
multi-hop network, this section presents preliminary results on
its performance. The results compare network operations in
the 868 MHz and the 2.4 GHz frequency bands and show the
advantages of intelligent scheduling of transmission cells in
6TiSCH in terms of packet delay. The experiments are con-
ducted using the open-source 6TiSCH simulator1developed
by the 6TiSCH community. The simulator natively supports
TSCH mode of the IEEE802.15.4e MAC layer and the RPL
routing protocol. The simulation parameters are listed in
Table I.
1) Frequency band evaluation: Figure 7 shows the Re-
ceived Signal Strength Indicator (RSSI) value for both the
868 MHz and 2.4 GHz band as a function of the distance
between transmitter and receiver. The sensitivity levels rep-
resent the lowest input signal the receiver can decode with
an acceptable signal quality. We differentiate between the
actual and the practical sensitivity levels. The former one
is the one advertised in the data sheets of the radio chips
and should be regarded as a theoretical optimal value. The
latter is a value measured by the manufacturer (i.e., Texas
Instruments2) in a practical setting and thus represents a
more realistic estimate. The results clearly show that operation
under 868 MHz supports a higher quality signal over a longer
distance, i.e., up to 700 m compared to 90 m in the 2.4 GHz
band (when comparing to practical sensitivity levels).
When looking at the total energy consumption, as shown
in Figure 8, the results for both bands are very similar. While
the absolute power consumption values for transmitting and
receiving with the CC1200 radio chip in 868 MHz are in
16TiSCH simulator: https://bitbucket.org/6tisch/simulator/src
2Texas Instruments Range Estimation Excel Sheet: http://e2e.ti.com/cfs-
file/ key/communityserver-discussions-components-files/156/Range-
Estimation-for-Indoor-and-Outdoor-Rev1 5F00 17.xlsm
Fig. 7: RSSI value comparison between 868 MHz and 2.4 GHz
IEEE 802.15.4g operation in function of the distance between
two nodes.
Fig. 8: Power consumption comparison between 868 MHz and
2.4 GHz IEEE 802.15.4g operation for a different number of
hops.
theory higher, the RSSI of the links between hops is also better
at this lower frequency. Better RSSI values result in a better
packet delivery ratio which leads to less time slots needed to
transmit each packet. This results in slightly better to equal
total energy consumption in the 868 MHz band as compared
to 2.4 GHz.
2) Scheduling function evaluation: To evaluate how the
intelligent allocation of time slots in a TSCH schedule affects
performance, both latency and throughput are compared for
SF0 and eLLSF in Figures 9 and 10. As expected, the through-
put for both scheduling functions is very similar and gradually
decreases with increasing number of hops. This decrease is due
Fig. 9: Throughput comparison for a different number of hops
between SF0 and eLLSF.
Fig. 10: Latency comparison for a different number of hops
between SF0 and eLLSF.
to the higher number of time slots being reserved when there
are more hops, which leads to a higher collision rate resulting
in a lower throughput. The latency results clearly show that
eLLSF outperforms SF0: up to a decrease of 27% in latency
when considering 5 hops. This is due to the daisy-chaining of
receive and transmit time slots that allows faster forwarding
of packets towards the sink, while SF0 allocates its time slots
randomly.
VI. CONCLUSION
In this paper, we present a novel IoT platform that sup-
ports reliable and guaranteed data dissemination and anal-
ysis, as well as actuator control, in dynamic and challeng-
ing infrastructure-less environments. The architecture, a case
study based on real-time monitoring and training adaptation
for cyclists and an initial hard- and software prototype are
described in detail. Evaluation results based on the presented
prototype show the promising abilities of this new multi-
hop-based IoT platform: (i) there is no additional power
consumption overhead when operating in the 868 MHz band,
(ii) the 868 MHz band increases single-hop distance from less
than 100 up to 700 m, and (iii) the latency is significantly
decreased when intelligently allocating time slots.
In the future, additional research will be done to develop
a disruption-tolerant, dynamic variant of RPL, operating on
top of IEEE 802.15.4g, which can deal with temporarily com-
munication disruptions and a continuously-changing network
topology. Also, an even more advanced TSCH scheduling
function needs to be developed that further decreases the
latency while taking the highly dynamic topology complexity
of the considered scenarios into account.
ACKNOWLEDGMENT
The authors would like to thank Bruno Van de Velde for his
valuable help in modeling the network energy consumption.
Part of this research was funded by the Flemish FWO SBO
S004017N IDEAL-IoT (Intelligent DEnse And Longe range
IoT networks) project, and by the ICON project CONAMO.
CONAMO is a project realized in collaboration with imec,
with project support from VLAIO (Flanders Innovation and
Entrepreneurship). Project partners are imec, Rombit, Energy
Lab and VRT.
REFERENCES
[1] J. Lanza, L. S´
anchez, V. Guti´
errez, J. A. Galache, J. R. Santana, P. Sotres,
and L. Mu˜
noz, “Smart city services over a future internet platform based
on internet of things and cloud: The smart parking case,” Energies,
vol. 9, no. 9, p. 719, 2016.
[2] S. Latre, P. Leroux, T. Coenen, B. Braem, P. Ballon, and P. Demeester,
“City of things: An integrated and multi-technology testbed for IoT
smart city experiments,” in Smart Cities Conference (ISC2), 2016 IEEE
International. IEEE, 2016, pp. 1–8.
[3] J. Jin, J. Gubbi, S. Marusic, and M. Palaniswami, “An information
framework for creating a smart city through internet of things,” IEEE
Internet of Things Journal, vol. 1, no. 2, pp. 112–121, April 2014.
[4] C. Pham, A. Rahim, and P. Cousin, “Low-cost, long-range open IoT for
smarter rural African villages,” in Smart Cities Conference (ISC2), 2016
IEEE International. IEEE, 2016, pp. 1–6.
[5] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-range
communications in unlicensed bands: The rising stars in the IoT and
smart city scenarios,” IEEE Wireless Communications, vol. 23, no. 5,
pp. 60–67, 2016.
[6] S. Aust and T. Ito, “Sub 1GHz wireless LAN deployment scenarios and
design implications in rural areas,” in GLOBECOM Workshops, 2011
IEEE. IEEE, 2011, pp. 1045–1049.
[7] J. Gozalvez, “New 3GPP standard for IoT [mobile radio],” IEEE
Vehicular Technology Magazine, vol. 11, no. 1, pp. 14–20, March 2016.
[8] I. Foche-P ´
erez, J. Sim´
o-Reigadas, I. Prieto-Egido, E. Morgado, and
A. Mart´
ınez-Fern´
andez, “A dual IEEE 802.11 and IEEE 802.15–4
network architecture for energy-efficient communications with low-
demanding applications,” Ad Hoc Networks, vol. 37, pp. 337–353, 2016.
[9] Z. Zaidi and K.-c. Lan, “Wireless multihop backhauls for rural areas: A
preliminary study,” PloS one, vol. 12, no. 4, p. e0175358, 2017.
[10] J. Simo-Reigadas, E. Municio, E. Morgado, E. M. Castro, A. Martinez,
L. F. Solorzano, and I. Prieto-Egido, “Sharing low-cost wireless infras-
tructures with telecommunications operators to bring 3G services to rural
communities,” Computer Networks, vol. 93, pp. 245–259, 2015.
[11] M. Weyn, G. Ergeerts, R. Berkvens, B. Wojciechowski, and Y. Tabakov,
“DASH7 alliance protocol 1.0: Low-power, mid-range sensor and actua-
tor communication,” in Standards for Communications and Networking
(CSCN), 2015 IEEE Conference on. IEEE, 2015, pp. 54–59.
[12] P. Kinney et al., “Zigbee technology: Wireless control that simply
works,” in Communications design conference, vol. 2, 2003, pp. 1–7.
[13] P. Thubert, “An Architecture for IPv6 over the TSCH mode of IEEE
802.15.4,” Internet Engineering Task Force, Internet-Draft draft-ietf-
6tisch-architecture-11, Jan. 2017, work in Progress. [Online]. Available:
https://datatracker.ietf.org/doc/html/draft-ietf-6tisch-architecture-11
[14] Bluetooth Core Specification, The Bluetooth Special Interest Group,
2016.
[15] “This is ANT: the Wireless Network Solution,”
https://www.thisisant.com/, Accessed: 2017-04-21.
[16] C. Perera, A. Zaslavsky, P. Christen, and D. Georgakopoulos, “Context
aware computing for the internet of things: A survey,” IEEE Communi-
cations Surveys and Tutorials, vol. 16, pp. 414–454, 2014.
[17] P. Barnaghi, W. Wang, C. Henson, and K. Taylor, “Semantics for the
internet of things: Early progress and back to the future,” Int. J. Semant.
Web Inf. Syst., vol. 8, pp. 1–21, Jan. 2012.
[18] J. Mineraud, O. Mazhelis, X. Su, and S. Tarkoma, “A gap analysis of
internet-of-things platforms,” Computer Communications, vol. 89–90,
pp. 5–16, 2016.
[19] S. De, T. Elsaleh, P. Barnaghi, and S. Meissner, “An internet of
things platform for real-world and digital objects,” Scalable Computing:
Practice and Experience, vol. 13, no. 1, pp. 45–58, 2012.
[20] J. Calbimonte, S. Sarni, J. Eberle, and K. Aberer, “XGSN: An open-
source semantic sensing middleware for the web of things,” in Proc. of
the 7th International workshop on Semantic Sensor Networks (SSN) at
the International Semantic Web Conference, 2014, p. 51.
[21] M. Ali, N. Ono, M. Kaysar, K. Grin, and A. Mileo, “A semantic
processing framework for IoT-enabled communication systems,” in
Proc. of the International Semantic Web Conference, 2015, pp. 241–
258.
[22] Indra, “SOFIA2: The first middleware for smart applications,”
http://sofia2.com/home en.html, 2017.
[23] D. Dujovne, L. A. Grieco, M. R. Palattella, and N. Accettura, “6TiSCH
6top Scheduling Function Zero (SF0),” Internet Engineering Task
Force, Internet-Draft draft-ietf-6tisch-6top-sf0-03, Mar. 2017, work in
Progress. [Online]. Available: https://datatracker.ietf.org/doc/html/draft-
ietf-6tisch-6top-sf0-03
[24] T. Chang, T. Watteyne, Q. Wang, and X. Vilajosana, “LLSF: Low
latency scheduling function for 6TiSCH networks,” in 2016 International
Conference on Distributed Computing in Sensor Systems (DCOSS), May
2016, pp. 93–95.
[25] A. Brandt, J. Vasseur, J. Hui, K. Pister, P. Thubert, P. Levis, R. Struik,
R. Kelsey, T. H. Clausen, and T. Winter, “RPL: IPv6 Routing Protocol
for Low-Power and Lossy Networks,” RFC 6550, Mar. 2012. [Online].
Available: https://rfc-editor.org/rfc/rfc6550.txt
[26] H. Stuckenschmidt, S. Ceri, E. D. Valle, and F. van Harmelen, “Towards
expressive stream reasoning,” in Proc. of Dagstuhl seminar on semantic
aspects of sensor networks. Schloss Dagstuhl - Leibniz-Zentrum fuer
Informatik, Germany, 2010, pp. 1–14.
[27] E. Della Valle, S. Ceri, F. Harmelen, and D. Fensel, “It’s a streaming
world! Reasoning upon rapidly changing information,” IEEE Intelligent
Systems, vol. 24, no. 6, pp. 83–89, 2009.
[28] A. Margara, J. Urbani, F. van Harmelen, and H. Bal, “Streaming the
web: Reasoning over dynamic data,” Journal of Web Semantics, vol. 25,
pp. 24–44, 2014.
[29] X. Su, E. Gilman, P. Wetz, J. Riekki, Y. Zuo, and T. Lepp¨
anen, “Stream
reasoning for the internet of things: Challenges and gap analysis,” in
Proc. of the 6th International Conference on Web Intelligence, Mining
and Semantics (WIMS). ACM, June 2016.
[30] X. Vilajosana, P. Tuset, T. Watteyne, and K. Pister, “Openmote: open-
source prototyping platform for the industrial IoT,” in International
Conference on Ad Hoc Networks. Springer, 2015, pp. 211–222.
[31] Powerful Wireless Microcontroller System-On-Chip for 2.4-GHz IEEE
802.15.4, 6LoWPAN, and ZigBee Applications, Texas Instruments, 12
2012, rev. April 2015.
[32] “nrf52 dk: Development kit for nrf52832,”
https://www.nordicsemi.com/eng/Products/Bluetooth-low-
energy/nRF52-DK, Nordic Semiconductor, Accessed: 2017-04-21.