Content uploaded by Gilles Privat
Author content
All content in this area was uploaded by Gilles Privat
Content may be subject to copyright.
Digiworld Economic Journal, no. 87, 3rd Q. 2012, p. 1. www.comstrat.org
Extending the Internet of Things
Gilles PRIVAT
Orange Labs
Abstract: This article proposes a theoretical and practical extension of the IoT, taking in
all "things" that can be sensed by sensors, without requiring them to be fitted with a tag or
a digital network interface. These physical entities, whatever they may be (legacy
appliances, passive items, subsets of physical space), become nodes of a broader
network, extending the internet of sensor/actuator devices. We explain how such an
evolution for environment-to-information interfaces draws upon a similar, long-standing
evolution of human-to-information interfaces. Multisensor acquisition of physical context
supports this extended IoT, bypassing the need for network-ready identification of target
entities. We describe a three-layer reference architecture for an infrastructure supporting
the integration of applications into the extended IoT. We show on a few examples how this
can expand IoT applications and endow them with features of robustness, scalability and
self-configurability.
Key words: Internet of Things, directed graph, physical context, multisensor data fusion,
pattern recognition.
Introduction
Beyond the internet of devices
Most mainstream visions of the "Internet of Things" come down to
extending the range of devices that may become connected to networks,
moving from Wifi or cellular to RFID, Zigbee or their equivalents. The oft-
repeated rationale is straightforward: there are trillions of such low-end
devices out there, waiting to get connected, when billions of humans and
their regular devices already are. If a new variant of Metcalfe's law
1
applies,
the promise of these "things to things" connections appears almost limitless.
1
According to which the value of a network increases in proportion to the square of the number
of nodes connected to it. A variant with lower exponent could possibly be derived for networks
of non-peer entities, i.e. nodes without full configuration and bidirectional communication
capabilities, as is the case for most “things” in the original or extended IoT.
2 No. 87, 3rd Q. 2012
Under such earlier catchphrases as "smart devices", "communicating
objects", "pervasive networking" or M2M, it is no surprise that the telecom
industry had been embracing this evolution as a legitimate extension of its
territory, well before the "Internet of Things" gained currency as the new
buzzword of choice.
Yet, compelling as it may seem, this "things-to-things" vision misses the
crux of the evolution: this is not only a quantitative extension of existing
person-to-person networks, it is a genuine quantum leap. By connecting
"things" that are deeply embedded in the physical environment, ICT systems
become strongly coupled with all kinds of physical systems. This opens up
entire new domains that were so far entirely outside the reach of ICT, or for
which information systems were disconnected from the corresponding
physical plant/system/process, requiring manual configuration and manual
data entry to couple the two.
Moreover, as viewed from the confines of the telecom industry, these
early attempts at redefining communication beyond person-to-person have
created some confusion because the distinction between the different
categories of new things/devices that became attached to networks was not
always made clear, especially when mobile phones and their avatars were
added to the mix. Devices for which sensors and actuators are used
exclusively to support regular human interfaces
2
should not be counted in
for the IoT proper. The devices that make up the Internet of Things in its
mainstream, yet strict sense, are networked sensors, actuators, devices
equipped with sensors and/or actuators, or more generally networked
devices that are not primarily ICT devices, equipped with physical interaction
capabilities that correspond to their primary function. This Internet of devices
comprises things/devices that are, in a proper sense, "embedded" in their
physical environment, adding information processing and transmission
capabilities to this environment
3
.
In this view, the outer border of the digital network is still the sensor or
actuator itself, beyond which is the physical world. The revolution of
pervasive networking that lead to the multiplication of these connected
sensors and actuators (PRIVAT, 2006) afforded an order-of-magnitude
2
Classical IT interface devices and telecom terminals use sensors for information input from
their human users and actuators for information rendering to their users: this is different from
using sensors to capture data from the environment and actuators to modify this environment.
3
This may in fact include regular ICT devices such as smartphones if they are used for
distributed network sensing rather than only information input/output from their users.
Gilles PRIVAT 3
enlargement in bandwidth between networks and the physical world. Yet, for
all their transformative role, this current generation of sensors and actuators
do not correspond to the ultimate possible displacement of the network
border. How this border may shift further to include real-world "things" is the
next stage of the evolution that we intend to describe.
Drawing upon human interfaces
Most grand schemes devised for the classical supply chain management
applications of the Internet of Things, such as the EPCglobal Network
4
, the
uIDCenter
5
, or "Internet 0" (GERSHENFELD, KRIKORIAN & COHEN,
2004) attach a universally unique, network-ready digital identity to these
physical things, be it their General ID, ucode or IP address. This amounts to
digitizing these "analog" things, or to shifting the border between the digital
and physical worlds further towards the digital.
In the latest evolutions of human interfaces
6
, this border has been
moving in exactly the opposite direction, which amounts to making the
digital/information/virtual world appear more like the analog/human/physical
world. Human interfaces are designed so as not to force human users to
meet the digital information world on its own terms, they should come closer
to an interaction between humans than to an in interaction between
programs or networked entities. The entire agenda of ambient intelligence
(STREITZ & PRIVAT, 2009) and such ideas as "perceptual interfaces"
(REEVES & NASS, 2000) or “tangible interfaces” (ULLMER & ISHII, 2000)
bear witness to this. The difference between data input through a keyboard
and command line interface and input through a software personal assistant
with voice recognition should make this clear. A less obvious example is the
replacement of clicking on a menu item by the grasping of a tangible
interface object that physically impersonates the same digital entity.
Obviously, the interface is moving much further into the analog world in the
latter case, and it requires more sophisticated sensing and perception
capabilities on the part of the system.
4
http://www.epcglobalinc.org
5
http://www.uidcenter.org
6
This does not refer to virtual reality, which could be considered as a counter-trend (or a mere
extrapolation of past trends).
4 No. 87, 3rd Q. 2012
These opposite evolutions have each been advocated for valid and
widely accepted reasons in their own right. As for human interfaces,
convenience and ease of use are not the only reasons for the un-digitization
trend: robustness, graceful degradation and reliability could be
complementary reasons, though this is not yet obvious with, e.g. the
replacement of keyboard text input by error-prone speech recognition. It
would be clearer if we could replace a password input by some 100%
foolproof face or palm recognition software. For supply chain management
applications that digitize everything from cattle to laundry detergent boxes,
attaching universal identifiers appears mandated by the need to scale up the
efficiency of digital data management to the physical world.
The main thesis of this article is that we can and should apply ideas
drawn from the domain of human interfaces to networks of physical things:
communication with and between these need not be forcefully digitized, it
may retain the specific properties of the physical world in which these things
belong, and gain new benefits from this. We explain in the following how
sensor networks can be extended and consolidated into an infrastructure
that supports un-digitized "thing-computer interfaces" inspired from ambient
and context-aware human-computer interfaces.
Related work
A few European projects, among which IoT-A
7
and FI-ware
8
(DE,
ELSALEH, BARNAGHIP & MEISSNER, 2012), propose a reference
architecture for the Internet of Things that highlights the distinction between
networked devices and real-world things (entities), in a sense close to the
one proposed here. This distinction is matched to different root categories in
the ontologies that support the relevant semantic models. These models are
domain-specific for physical entities of the environment, and intended to be
shared with applications. Our work goes further in extending the graph
network model to these things and, crucially, the automatic discovery and
configuration mechanisms that make it possible to integrate them in the
network as if they were regular devices.
7
www.iot-a.eu
8
www.fi-ware.eu
Gilles PRIVAT 5
Network and distributed software infrastructures with capabilities for
spontaneous "zero-configuration" discovery and integration of new devices
are a clear inspiration for the solution presented here. These infrastructures
differ widely in the levels of interoperability they include in their scope. At the
lowest end is the spontaneous configuration of network-layer addresses with
mechanisms such as the automatic assignment of private ("link-local") IP
addresses. At a higher level, distributed Service Oriented Architectures
9
view devices through the description of the services they support. These
services may get discovered and advertised once the devices are
connected. A number of research and prototype solutions (SONG,
CARDENAS & MASUOKA, 2010) have been proposed to extend these
mechanisms to the semantic level. This would relax the requirement for prior
definition of the corresponding services in the format required by the
standard, for them to be discovered. Yet it does still require interoperability
at the lower levels. However ambitious and high-level they may be, all these
infrastructures are strictly limited to classical networked devices. They
require that the target devices be natively endowed with network-ready
interfaces or tags, and that the corresponding interfaces comply with
required standards at all appropriate levels. The most constraining
requirement is that devices need to be fully known beforehand as instances
of extremely specific types
10
for these mechanisms to work. Identifying the
device to a general category (such as "printer" or "display") is very far from
sufficient for a particular instance of this category to be integrated in a
regular service-oriented architecture.
It is in these regards that the proposed approach differs most radically
from existing infrastructures. It requires neither equipping devices with
standard interfaces, nor pigeonholing them to a specific type. Devices can
be identified by approximation to a very generic model, and the system
should be able to integrate them on the basis of this minimal information.
Under the general "ambient intelligence" agenda (STREITZ & PRIVAT,
2009) the "smart space" research strand has targeted the physical
environment as a primary basis for context grounding. In this view, devices
are just transparent intermediaries and the environment itself may become
an interface for human interaction. In placing the focus "beyond devices",
9
Such as UPnP (www.upnp.org)
or DPWS (http://docs.oasis-open.org/ws-dd/ns/dpws/2009/01).
10
This type information includes at least to the manufacturer, the brand/make, the commercial
name/model, the year of issue/version number.
6 No. 87, 3rd Q. 2012
this has much in common with the approach we propose here, which can be
extended to relevant subsets of the environment as target physical entities.
These "space entities" can be represented and modeled in a way very
similar to "graspable" entities, using the same infrastructure to interface with
applications that use high-level context information about the environment.
The extended IoT of sense-able/actionable things:
reference model
Extending the network to "sense-able" things
We start from a narrow mainstream view of the Internet of Things as a
network of sensor-equipped devices and examine how to make the external
interfaces of this network more "thing-friendly", as if they were perceptual
human interfaces. Instead of constraining things to adapt to the network, we
make it possible for the network to adapt to them. Instead of enforcing their
uniform digitization, we try to take them as what they are, analog physical
things.
Let’s for the time being take the example of a single sensor, e.g. a
camera, and suppose we have a "thing" recognition and monitoring software
analyzing the data acquired by this camera. Assuming this, we can consider
that every individual thing or physical entity within the field of view of this
camera becomes ipso facto a "networked thing", provided it can be
recognized and monitored by this software coupled to the camera. This
means it can have a presence on the network, without requiring an RFID tag
or even a digital optical code (such as a 1D or 2D barcode) for this. We
proposed in earlier work (PRIVAT, 2012) to extend the notion of
"phenotropics"
11
as a broad conceptual basis for the use of thing
recognition and monitoring through sensors as a new type of analog network
interface. We will not be delving further into these theoretical aspects here,
focusing instead on practical and implementation concerns.
In this view, the range of things that may become indirectly connected to
the network can extend much further than sensor devices themselves, to all
11
A concept and noun originally forged by VR pioneer Jaron Lanier, who viewed it first as a
way to make the internal interfaces of information systems more robust and adaptable.
Gilles PRIVAT 7
things that are individually identifiable by a sensor This extension of
networks to new nodes "beyond sensors" is represented in figure 1 by a set
of directed "sensor to thing" links. It is not a routine incremental and
quantitative extension, such as results from adding a new wire-line or radio-
based protocol. It is a qualitative leap that makes it possible to integrate all
discrete or bulk "stuff", as it is, without requiring any digital identity or
network interface whatsoever, and without adhering to any kind of
predefined standard, at any level, for this connection. There is no prior
barrier to the integration of new things. It is by nature universal as it requires
no prior standardization of any kind of digital code or interface.
Extending the network to "actionable" things
As the counterparts of sensors, actuators transduce numerical variables
into physical ones. They enact modifications of the physical environment and
the effects of these are either sensed directly by sensors, or indirectly,
through passive things which are modified by the actuators. These new
physical "actuator-to-thing" links complement the "sensor-to-thing" links.
Together they make up a directed graph, or virtual network that we have
proposed to call a stigmergic network (PRIVAT, 2012), by reference to a
biology-inspired concept of complex systems theory
12
. This "actuation
graph" is again overlaid upon the wire-line/wireless data network through
which the corresponding sensors and actuators receive or transmit their
respective numeric data (figure 1).
What we integrate in the network here is not new nodes, but new links
that "close the loops" of sensor-actuator networks in a way that does not use
the modalities of classical networks and complements them. New qualitative
system-wide properties can be analyzed in this double coupling of sensors
and actuators: through the network and through their common physical
environment. This coupling may result in both extremely useful and
potentially undesirable or harmful effects, making this a key challenge for
future research. We will not elaborate here on these system-theoretic
aspects.
12
Stigmergy, a concept originally proposed by Pierre-Paul Grassé from his observations of
social insects, refers to indirect communication between agents by the modification of a shared
physical environment.
8 No. 87, 3rd Q. 2012
Figure 1 - Internet of Things extending the Internet of Devices
(extension links as dotted arcs)
Extended Internet of Things, or Web of Things?
If we called "Internet of Devices" the network of sensors and actuators
(devices that are attached to networks in a traditional sense), should we call
the extended network proposed above the "Internet of Things" proper, or the
web of things? Much as the early World-Wide-Web that we know of was a
virtual network of hyperlinked static HTML documents overlaid on top of the
Internet, the network of sensed things can be considered as a virtual
network extending an internet of devices and sensors, which is itself an
extension of the early Internet
13
. As a virtual network (a graph in
mathematical terms), this network of things comprises a far larger number of
nodes than an IP network ever will, just as the web has many more
documents than internet hosts. Another key difference is that, whereas the
graph representing an IP network is non-directed
14
, the graphs representing
13
Represented by the inner circle of “hosts” in figure 1.
14
This means its links (edges in graph theory terms) are bi-directional.
Gilles PRIVAT 9
either the classical document-centric web, or the extended Internet of
Things, are directed
15
.
Even though the so-called "web 2.0" has been more of a fuzzy
amalgamation of hype than a distinct technological evolution, no less an
authority than Vinton Cerf has claimed that the "web 3.0" would be the future
Internet of Things. The "web of things" is an alternative name that could be
used, hadn't this phrase already been proposed (GUINARD & TRIFA, 2009)
to capture the use of lightweight application-level web-based protocols for
the Internet of Things. To avoid confusion, we will keep to the phrase
"extended Internet of Things" (using "xIoT" for short) in the rest of this article.
Contextual interfaces and multi-sensor pattern recognition
The notion of bilateral "sensor to thing" links presented above is a simple
abstraction of the more diffuse, multilateral reality of context sensing that
should actually apply for the identification and monitoring of things in the
extended IoT. Here again, a very useful lesson can be drawn from the
evolution of human interfaces.
In the course of moving away from the digital border of networks, human
interfaces have become context-aware, which means they have been
evolving from a simple bilateral human to device relationship to become
mediated by the virtual and physical environment in which the interaction
takes place. To focus here on physical context, a context-aware interface
takes into account other sensor inputs than the primary, explicit user inputs,
like when the user’s location
16
is brought to bear to scope a request for
some local service, without the user having to specify it explicitly. In a
broader view, context-aware interfaces amount to using the entire physical
environment, rather than one single device, as an interface. This
environment becomes a smart, perceptual environment, where all sensors
15
This means its links (arcs in graph theory terms) are uni-directional. For the extended
internet of things only the extension links (things → sensors and actuators → things arcs) are
unidirectional , whereas all network links in the original web are unidirectional: this is related to
the difference between an extension graph, which is a superset of the original graph, whereas
an overlay graph like the web is in a different plane, its nodes being mapped to the nodes of the
underlying graph.
16
Acquired through a sensor-equivalent location-determination technology which, whatever it
is, can be considered as providing implicit secondary data complementing the user’s explicit
input.
10 No. 87, 3rd Q. 2012
are federated, acquiring low-level context data that is fused and interpreted
to become high-level context.
A similar notion of physical context should apply to the "things to
networks" interfaces, for them to become "less digital". All sensors of a given
environment can be brought to bear in order to "connect" things. We again
assume that we have available through the "Internet of Devices" a federation
of distributed networked sensors that make up a "smart environment". These
sensors are not dedicated to one application and they can all provide useful
context data about this environment. Networked "things" in this indirect,
extended sense may then comprise all "stuff" that can be sensed by pattern
recognition software operating on top of these federated sensors working
together, potentially overcoming their individual limitations as single-
modality
17
devices. Whether they are primary sensors, or provide only
complementary data, the identification and monitoring of "things" in this
environment will use them jointly, opportunistically, taking into account their
data inasmuch as it is relevant.
The kind of pattern recognition used to identify and monitor things on this
basis is very different from classical pattern recognition based on separate
modalities, such as used in computer vision or speech recognition. Pattern
recognition used in this way may rely on classical multi-sensor data fusion
(HALL & LLINAS, 1997), but when using very basic sensors such as passive
infrared, door contact or electrical sensors it is in fact much simpler than
when dealing with rich and complex signals such as video or audio. It
assumes primarily the detection of temporal coincidence of events from
different sensors (as potentially coming from the same physical entity) and
the application of simple filtering rules to these multi-sensor events. The
consolidation of these events will then depend on the corresponding model
of the originating physical entity. More concrete examples and descriptions
of this are provided in the following sections.
17
Modality is used here in a sense derived from its use in human computer interaction, but is
not limited to human sensory channels: a modality is a type of physical variable or phenomenon
that is measured or detected by a sensor, such as temperature, pressure, location, movement,
etc.
Gilles PRIVAT 11
From universally unique to contextual identification
Things that are connected to the extended IoT in the sense proposed
here need not have an explicit, pre-assigned and pre-registered universally
unique identification attached to them, as is normally a prerequisite for
mainstream IoT "things" with classical technologies such as RFID. It is in
fact sufficient for most applications to assume that things are implicitly
identified. This means that they are identified on a relative, non-universal
basis in a given context or scope. It is the knowledge of the context,
combined with the local identification, which can make the identification of
things global and universally unique if needed. The software proxy
(representative) of these things in the network will maintain this identification
in a way that can be used by applications. Human computer interfaces are
again evolving in the same direction, requiring a unique universal
identification from users only when it is needed, as most users will prefer a
contextual identification that preserves their privacy (such as e.g. a session
cookie for web-based applications) when it suffices.
The key difference with traditional network-based identification is similar
to what was already mentioned for network connection. Contextual
identification does not require prior standardization and shared knowledge of
a set of codes or protocols for exchanging this identification. It does not
require either a prior registration of the object in a database. The entity may
be identified on the basis of its physical features, as these are sensed by the
available sensors, associated with required context data as a complement.
The "recognition" of the entity and its association with a known category
relies on generic, publicly available knowledge, not on a matching with some
proprietary database. If not unique as an instance of this category, context
(such as location) comes in to complement the category-based
identification
18
.
Extended IoT: reference architecture
From the proposed conceptual model, we derive a layered
software/network architecture that can serve as a reference framework for
18
Universally unique identification à la EPC global may of course still be a requirement for
some applications such as supply chain management, but this does not mean it should be
enforced when it is not needed.
12 No. 87, 3rd Q. 2012
the implementation of ICT systems based on the extended IoT as defined in
this article. This architecture draws an analogy to the layering of both
standard networking models
19
and computer architectures, where the
underlying physical hardware is abstracted away, hidden under successively
more abstract and hardware-independent interfaces provided to applications
(figure 2).
Extended IoT framework
From the viewpoint of an application using the extended IoT, the problem
addressed can be stated in the most general possible way as follows: an ICT
system is set up to acquire data from a large-scale physical system and, if
need be, control it in return. Shared sensors and actuators are distributed as
monitoring and control points through this physical system. A set of entities,
subsystems of the overall physical system, are defined as the components
of the overall physical system that are relevant for being controlled and
monitored by the targeted application. These subsystems are distinct
physical entities. They are fully-fledged physical systems in their own right:
they will be the nodes of the extended IoT for this application. The sensors
and actuators are not target entities themselves. They are used just as
transparent intermediaries.
The ICT system will "shadow" each of these xIoT nodes individually
through matching software components (proxies) that will offer to
applications the required interfaces to the extended IoT in this environment.
The ICT system should have the capability to create and configure these
components automatically. This is required for the initial configuration stage
and when a change in the environment triggers a reconfiguration. The
configuration includes the automatic association with the entity proxy of the
interfaces to the subset of sensors and actuators that are used as
intermediaries for the monitoring and control of a given entity.
19
Such as the (7 layer) ISO or (4 layer) internet models.
Gilles PRIVAT 13
Abstraction of xIoT nodes
We choose to represent the target entities (xIoT nodes) through simple
discrete-state models transitioning on discrete events
20
. Their digitized
states, possibly complemented with relevant continuous-valued attributes,
are then stored as the state of the proxy of the xIoT node. These models
represent a simple yet adequate common denominator abstraction of reality
for many practical xIoT use cases. They provide, as the centre points of
discrete classification clusters, the necessary "anchor" for making sense of
multidimensional sensor data without resorting to complex pattern
recognition techniques.
Self-configuration and reconfiguration
Self-configuration makes it possible to identify and integrate
spontaneously and automatically new entities into the xIoT network. The
process goes through the following stages:
Detection of meaningful sensor event as corresponding to a new
entity
Creation of a generic entity proxy in the corresponding entity
abstraction layer
Association of this entity proxy with the sensors that provide data
about this entity
Update of the entity model from additional sensor data
Assignment to more specific entity category and more specific entity
model
Association with complementary sensors
From step 6, the configuration process may iterate in a loop including
steps 4 to 6 until the most specific available model is reached.
Re-configuration refers to the continually on-going adjustment of the
system to account for changes in the environment, such as removing or
20
This means their status is captured by a state vector which is a discrete-valued function of
time with discrete asynchronous (event-based) transitions, defined in a system-theoretic sense
as encapsulating the necessary and sufficient information to obtain the next states and outputs
of the system given its next inputs. Cf. CASSANDRAS & LAFORTUNE (2008) for an
introduction to discrete event systems.
14 No. 87, 3rd Q. 2012
adding a new sensor/actuator, moving an entity or removing it. It is triggered
by a mismatch between the sensor data and the entity model. From there, it
involves a backward traversal of the graph of entity models
21
until the data
matches again with the model. The configuration may then start over, just as
an initial configuration, going through the very same stages.
Entity group representation
Proxies representing individual physical entities are all distinct software
components at the same hierarchical level and do not contain one another,
even if the corresponding entities have such containment relationships: for
example, the proxy of a room will not contain the proxy of an appliance, even
if this appliance is inside the room. An additional separate layer of "entity
groups" is needed to represent such aggregation or containment relation
ships between physical entities, with a 1-to-n or n-to-1 mapping to the
physical entity layer. A virtual entity may thus link to several physical entities
or a physical entity may link to several virtual entities. The relationship
between these different layers is represented in figure 2.
Interface to applications
The interfaces that are exposed to applications from the proposed xIoT
architecture abstract away sensor and actuator data at a level corresponding
to the states and associated attributes of the entities, as defined above.
For monitoring an entity through the xIoT, an application can obtain the
instantaneous state of this entity as the discrete state of the corresponding
entity proxy, associated, if required, with complementary continuous-valued
attributes. This discrete state is estimated as a result of the fusion,
aggregation, consolidation and classification of data from sensors
associated with the entity.
For control purposes, an application can effect a change in the state of
an entity to another admissible state through the entity proxy that relays this
21
These models and their inheritance relationships make up a directed acyclic graph (DAG),
which results form merging the arborescences corresponding to different complementary
classification criteria. This DAG is traversed from the roots (the parentless; most generic
models) to the leaves (the childless, most specific models) in the initial configuration phase, and
traversed back from the leaves to the root when a reconfiguration is triggered.
Gilles PRIVAT 15
high-level state-change control order to low-level control data for the
associated actuators.
Figure 2 - Reference architecture for the extended Internet of Things
Application examples
Integration of legacy appliances in a home network
In the home environment, target entities are those that are relevant for
being monitored and controlled by applications such as energy
management, security/safety management or home automation, extending
the home area network beyond its regular perimeter of ICT devices and
state of the art home automation devices.
16 No. 87, 3rd Q. 2012
If we take home energy management as an example, examples of the
target physical entities would be:
- appliances and devices of all types, including all pieces of legacy
home equipment,
- rooms of the home,
- energy-relevant components of the home such as walls, windows.
For other applications, this might be extended to pieces of furniture, pets,
or the home occupants themselves. These mostly non-digital entities have to
be integrated in the home xIoT network in a way similar to what is done with
regular networked entities. This means they have to be identified and
matched to an existing model that can be specific or generic, exact or
approximate. State of the art devices would afford this integration through a
high-level SOA-like interface, but until they are widespread in the home
domain (which will take a long time because of the slow turnover of home
appliances), we still have to deal with legacy appliances whose only
available interface is that of their mains connection. This interface makes it
possible to identify these appliances through the characteristic features of
the patterns exhibited by their electric power consumption through an
electric power sensor (like e.g. an oven showing a steady plateau pattern
whereas a washing machine has characteristic peaks and troughs). This
electric current sensor will be the main sensor for mains-connected
appliances, possibly complemented by other sensors available in the
environment. When these appliances are identified and enrolled into the
extended home IoT in this way, it becomes possible to monitor and control
them as specific or semi-generic entities, even though this control is limited
to the mains interface. This is not equivalent to what can be done through a
state of the art data network (which would in principle make it possible to
remotely program the appliance, or at least change its mode of operation)
but it may still be sufficient for the purpose of monitoring and controlling it for
energy management (HU & PRIVAT, 2011).
Multiscale energy management in the Smart Grid
The smart grid can be considered as the result of adding an IoT layer on
top of the electrical grid. The smaller scales of the smart grid may involve the
decentralized management of semi-autonomous units such as home,
building or district "micro grids", where all kinds of electrical equipment and
energy-relevant physical entities can get integrated in a local energy
management system.
Gilles PRIVAT 17
Figure 3 - Extended IoT supporting multi-scale decentralized energy management
in the smart grid
As explained before for the home and building domains, these entities
are widely diverse and heterogeneous, adding to the mix of home/building
appliances the type of power equipment that gets connected to the
distribution network. This may be classical electrical engineering machinery
such as inverters and transformers, or the newer type for which the smart
grid is precisely intended, like distributed renewable energy resources (e.g.
18 No. 87, 3rd Q. 2012
wind turbines or PV panels). We advocate that the "extended IoT" as
proposed and described here is the proper approach to the design of the ICT
layer of the smart grid because, among other reasons, it is essential to the
viability of this extended and decentralized smart grid approach that the
integration of these entities does not require manual configuration. The
nested scales corresponding to home, building and district energy
management systems, together with examples of the corresponding entities,
are illustrated in figure 3 (HU & PRIVAT, 2012).
Conclusion
The proposed approach has been developed and validated so far in the
home domain, on the basis a complete range of home entities, from
electrical appliances to rooms. The home area network thus extended may
be considered, if used for energy management, as a home-scale smart grid,
intended to nest within larger scales of the smart grid using the same
extended IoT approach. The larger scales that we may deal with will
correspond to cities or city districts, for which smart grids are but one
application. Relevant target entities for smart cities might be as diverse as
lampposts, garbage containers, cars, pedestrians and building themselves.
Besides this, all classical applications of the Internet of Things and smart
environments could be revisited by applying this approach, relaxing the
requirements for state of the art network interfacing and digital identification
of physical entities, making them more open, "things-friendly" and ultimately,
we hope, more widely accessible and successful.
Gilles PRIVAT 19
References
CASSANDRAS, C. G. & LAFORTUNE, S. (2008): Introduction to Discrete Event
Systems, Springer.
DE S, ELSALEH T., BARNAGHIP, P. & MEISSNER S. (2012): "An Internet of Things
Platform for Real-World and Digital Objects", Scalable Computing: Practice and
Experience, Vol. 13, no. 1, pp. 45-57. http://www.scpe.org
GERSHENFELD, N., KRIKORIAN, D. & COHEN, R. (2004): "The Internet of Things",
Scientific American, October.
GUINARD, D. & TRIFA V. (2009): "Towards the web of things: Web mash-ups for
embedded devices", in Proceedings of WWW conference, Madrid, Spain
HALL, D. L. & LLINAS, J (1997): "An introduction to multisensor data fusion,"
Proceedings of the IEEE , vol. 85, no. 1, pp. 6-23, January.
HU, Z & PRIVAT, G.:
- (2011): "Iterative Model-based Identification of Building Components and
Appliances by Means of Sensor-Actuator Networks", Computers in Buildings 2011,
Workshop on data models for energy-efficient buildings, Sophia Antipolis, October
2012.
- (2012) "Representation and Self-Configuration of Physical Entities in Extended
Smart Grid Perimeter", IEEE Innovative Smart Grid Technologies (ISGT), Berlin,
October.
PRIVAT, G.:
- (2006): "From Smart Devices to Ambient Communication", Invited conference,
From RFID to the internet of Things, EU Workshop Brussels, March 3 2006.
ftp://ftp.cordis.europa.eu/pub/ist/docs/ka4/au_conf670306_privat_en.pdf
- (2012): "Phenotropic and Stigmergic webs, the new reach of networks", Universal
Access in the Information Society, Vol. 11, no. 3, pp. 323-335.
REEVES, B. & NASS, C. (2000): "Perceptual user interfaces: perceptual bandwidth",
Commun. ACM 43, 3 (March), 65-70.
SONG, Z., CARDENAS, A. & MASUOKA R. (2010): "Semantic Middleware for the
Internet of Things", Internet of Things 2010.
STREITZ, N. & PRIVAT G. (2009): "Ambient Intelligence", Chapter 60 of Universal
Access Handbook, Constantine Stephanidis, Editor, CRC Press.
http://www.crcpress.com/product/isbn/9780805862805
ULLMER, B. & ISHII, H. (2000): "Emerging Frameworks for Tangible user interfaces",
IBM Systems Journal 39(3&4): 915-931.
