Page 135 - ITU Kaleidoscope 2016

ICTs for a Sustainable World

such mesh. As now happens with applications made up of costs involved in case of ad-hoc sensing infrastructure, when
distributed components overlaid on the network, interacting not absolutely required.
by means of well-established protocols and paradigms, e.g. Dynamic behavior and chances for data reuse are also differ-
RESTful services, also things should network themselves entiating factors in evaluating MCS against WSN-like con-
and their corresponding agents, with no (direct) operator figurations. When dealing with MCS, the composition of de-
intervention (think M2M). Under such premise, MCS may vice populations, as well as the kind and quality of data pro-
just become a pattern under the IoT umbrella term, i.e. a spe- duced, where quality may be expressed both in terms of la-
cialization of the platform that an IoT would represent for tency and precision, may change in time due to mobility pat-
sensing-related, mobility-enabled, crowd-sourced use cases. terns, power requirements and communication subsystems,
This perspective should be particularly appreciated in light of including owner-mandated local preferences. In traditional
opportunistic developments, as attaching semantic descrip- WSNs the populations (and the kind of data produced) are
tion to things, endowed with distributed, event-triggered usually known a priori, thus managing quality and designing
logic, may really provide chances for nodes to discover each according to requirements is more straightforward. More-
other services and weave an even more powerful abstraction over, with regard to MCS, support for multiple concurrent
with respect to the Web of Things. applications is feasible, albeit subject to careful planning and
Although IoT can be mainly associated with the participa- design stages. When it comes to sensor networks instead, de-
tory pattern, some work on opportunistic IoT and sensing ployments are typically geared for a specific application, thus
environment is available in literature. For example, in [9] repurposing or resource sharing are rarely accounted for.
an opportunistic IoT framework is proposed, mainly extend- Still casting challenging constraints into opportunities, evo-
ing opportunistic networking towards participatory sensing, lutions in time of crowds’ shapes and composition may lead
enabling opportunistic information sharing among things to to fast dissemination of information, aiding in the pursuit of
also support mobile social networking. Similarly, oppor- (global) optimization objectives. Moreover we care about
tunistic mobile networking is the topic of [10], mainly focus- actual MCS usefulness, for contributors themselves too, thus
ing on low level data forwarding issues through a framework any solution can’t be considered absolutely all-round if lack-
able to support and optimize opportunistic sensing. The best ing built-in feedback and assisted guidance. Even with re-
way to exploit the seemingly limitless IoT potential for sens- gard to data itself, both in terms of raw format and appli-
ing at a global scale is by coupling opportunistic and par- cation requirements, sensor readings may not be suitable for
ticipatory application patterns for mobile crowds with the direct consumption. In this sense specific local analytics may
infrastructure underneath. The proposed solution is aimed be performed on-board, to (pre)process raw data, in order to
at horizontally enabling things for massive development and obtain intermediate results ready for transmission and further
deployment of any sensing task and based on a novel appli- processing.
cation of Cloud computing technologies to IoT, as described Even not taking into account the scalability issues any kind
in the following. of backend may be subject to when dealing with huge raw
As we pointed out already, not only should we optimize any data uploads, on-board resources in mobile devices are on
MCS despite shaky grounds, but we aim to turn most short- the rise and already remarkable in many cases, thus it seems
comings to our advantage. More specifically, uneven global natural to tap into such potential to shift in part the burden of
connectivity when coupled with dense, mesh-like topologies computations toward the edge. Moreover, a renown tradeoff
lends itself naturally to be addressed by means of routing pro- in conventional mote-class WSNs lies in leveraging (local)
tocols as those designed for MANETs. This leads to avoid- computation to save on the amount of energy and bandwidth,
ing otherwise likely network congestion and servers’ over- as needed to transmit preprocessed data in comparison to raw
load while also making room for any kind of crowd-local, readings. Last but not least, delay-sensitive applications may
opportunistic and cooperative behavior. benefit from such savings, if the overhead of local processing
is negligible when compared to transmission-induced laten-
In particular when compared to the building blocks of
cies.
MANETs par excellence such as, e.g., WSNs, nowadays
consumer-class mobile devices feature plenty of comput-
ing, storage and communication capabilities, making these 3. A SERVICE-ORIENTED MCS
platforms many orders of magnitude more advanced than INFRASTRUCTURE
mote-like ones, also in terms of sensing, considering that
mobiles are usually multi-modal by default, thus enabling a The Stack4Things framework [1, 11], also referred to as S4T,
wider range of applications. Moreover, the impact of massive is our effort to pursue a new approach for the management of
deployments in the field can never be overestimated, as these smart objects and things in the IoT scenario, following an
devices are available by the billions, always-on and mostly on-demand, service oriented provisioning model. This shifts
online, a consistent slice of people’s daily activities and re- the IoT paradigm towards the Cloud one, merging benefits
curring habits. Leveraging such huge populations means of both: on the one hand providing control and management
building potentially instant-on large-scale applications while capabilities to IoT (sensing and actuation) resources, on the
at the same time lowering expenses and time-to-market, also other enriching the Cloud paradigm with pervasive I/O capa-
avoiding altogether the long setup times and high upfront bilities to directly interact with the environment. Altogether

– 117 –